I liked how everything seemed intentionally rectangular. Perhaps made it easier to organize things mentally. It was also nice having full menus at the top of windows.
There was also a big technical reason for that: graphics subsystems sucked back then. Hardware acceleration ranged from non-existent to primitive and RAM was tight. If you have two objects (window, control, etc.) displayed and they're rectangular, you can update one of them without needing to involve the other one at all. Since your entire display was a single buffer in memory, each application would draw directly in its chunk of memory. When you did something which made things move you could use simple array math to move the previously-painted block of memory to the new location or, if you had a really fancy graphics card, call a BitBlt operation which would move x1,y2-x2,y2 to the new location. Then you had to tell whatever was now visible in the source location to repaint that region.
Adding things like transparency or rounded corners breaks a lot of that — you need to know what's underneath to update the display — and that wasn't fast enough to use until you had enough RAM to store each object independently so you don't need to ask each application to repaint as things move around. Compositing window managers started early — I believe the Amiga had that — but the first mainstream OS was when Mac OS X shipped Quartz and it had performance drawbacks (i.e. tolerable but not great window dragging performance) until they shipped Quartz Extreme which used OpenGL to do the hard work. Microsoft caught up about 5 years later with Windows Vista, in part because they had more users to support with inadequate hardware.
Apps could draw directly to DC, but it flickered a lot, and everyone quickly ended up with CreateCompatible{Bitmap,DC} and blitting into BeginPaint’s DC after drawing. After all, at win95 you had 4 to 8 megabytes or RAM - you definitely had the space. Not sure when it happened, but I can remember fighting with flicker circa 97. Also, rectangular controls were what you had in standard winapi, which was based on HWNDs, which were rectangular-only. I don’t think that unroundness was really forced by hardware constraints (transparency - yes), it was just a default simplest flavor of time. (And not that I like round/flat/text-only controls. What we called “flat” back then is pretty 3d-bordered in modern terms, and what we called “round”, today is border-radius:3px)
4-8MB wasn’t all that much if you were running more than one app at >640x480x256. It was certainly possible but fancy visual effects were going to exclude a lot of low-end hardware, which is what a LOT of mainstream users bought since hardware cost a good but more back then and the value proposition was less established.
I agree that roundness wasn’t prohibitive in general. The rectangular nature of the APIs encouraged it but you certainly saw it in applications. What I was thinking about most was complex usage where there was overlap outside of the active context (e.g. dragging a window with the expectation of a lower layer showing around the curve, rather than either not having that or being able to assume a solid fill color) or interfering with hardware accelerated painting or later video decoding which only worked in rectangular regions.
OpenBSD has canfield (solitaire), cribbage, [go]fish, mille [bornes], and even monop[oly] and bcd (format input as punch cards) out of the box (and way more! this is just a fraction of intro(6): http://man.openbsd.org/man6/intro.6). Unfortunately I’m not aware of any GNU distros that do, but you can get them with a simple `sudo apt install bsdgames` on Debian (also emacs has tons of games of course https://www.emacswiki.org/emacs/CategoryGames).
Debian in the WSL and Docker are plenty enough for work. I spent too many years installing/testing distros, and I'm tired of having to maintain my system (I was a Gentoo user for 2 years, and I built an LFS that I used for 8 months).
Even Debian, if you're not careful, can become a mess. Windows is dumb enough that you don't need to be careful.
Chocolatey was the last thing I missed from Linux/Unix distros.
And I can still code my own card game if I really want to :)
EDIT: seeing a sibling talking about slackware, it seems I missed the joke ^^
The contrast and distinction between the different UI elements is immediately obvious. No mistaking a button for static text, and vice-versa. Also, no hidden scrollbars or other auto-hiding UI elements that are difficult to discover and irritating to use even once you know about them.
Even if it looks 'ugly' by today's standard those were good days UI/UX wise and things kept on improving. UIs were consistent and easy to use by mouse or keyboard and in my opinion things started going downhill ever since. I think good UI/UX will of course be rediscovered and the old ideas will be recycled into something new at some point.
To offer a counterpoint - back then everything in an interface had to be painfully obvious and usable in a myriad of ways because a significant portion of users were computer-illiterate first timers.
As it is now, and forever will be, I think the balance between usability and ease of development will be those 20% of effort to make your app palatable to 80% of the users (barring some exceptions by people with underdeveloped moneymaking sense whose self-description might involve the word "artisanal".)
Yes, you have a point. When IPads/Iphones first came out they were so usable that old people who never touched a computer or toddlers could use them without much direction, they were quite intuitive. They've messed up since then though, they've gone into a dubious direction. And the rest of the industry followed good and bad apple UI/UX decisions, some of them even poorly and we have a confusing mess, confusing and headache inducing at times even for the computer literate.
I find touch interfaces have infantilized computing, turning everything into scrolling and big, fat, ad-looking buttons, and there is no way to really become skilled at them.
- Easy keyboard shortcuts with underlined keys in menus (Alt-F Alt-O!), and reconfigurable in general,
- F1 Help working everywhere,
- Unified keyboard shortcuts: F2 to rename, F5 to refresh (except in Lotus which closes entirely),
- Reconfigurable an draggable toolbars!
- Icons (those don’t exist in the web)
And so on. Treeviews! Lists with sortable and reorderable headers! They were awesome. It was the ultimate UI. Then they flattened everything and, worse than everything, came the time where treeviews disappeared and were replaced with infinite list sorted by “last document opened” like in Google Drive... That is what I regret most: Old documents being buried behind an infinite scroll, you can never be sure that you see and sort everything.
Remember selecting more than one of something, and then being able to drag-and-drop it somewhere, or right click it and perform actions on the whole list?
Remember tables with sortable column headers, in every app?
Oh man. The web sucks so hard, and the kids working on it now are so used to it they don't even know how hard they've been fucked over. To think that in 2021 people still have to think about how to wire up a data table to some backend, thousands of people writing their own version of that every year....
Imagine. The computer giving you a list of the actions you can do, written in a language you already understand, organized hierarchically by the kind of thing you are operating on. And as a bonus, most of them have keyboard shortcuts that are/were immediately spelled out directly on the menu!
The other day, I was using the Twitter web interface, and I realised I'd accidentally clicked the "Like" button on a tweet as I was going back to the previous page. Oh no, I haven't liked a bad tweet, have I? Of course, I go forward, and the page has completely changed, showing me a bunch of stuff that wasn't there the last time I saw it. Great, now I need to navigate the UI. I scan the interface for some sort of "likes list". It's not in the sidebar. It's not in the "more" in the sidebar. It's not even in "lists".
So I open Tweetbot, search the menu bar for "like", find it in "Tabs → Likes", and perform the action I want in seconds. Web applications can never compare.
It helped that Microsoft put out a design manual laying out their standards: "The Windows Interface Guidelines for Software Design: An Application Design Guide"
IMO things started going downhill the moment laptop manufacturers started putting touchscreens into laptops. No one uses them with any kind of seriousness, but apparently every single UI designer now designs with them in mind for some bizarre reason.
Yes, mixing the 2 paradigms was a disaster. Windows 10 got a lot of pushback when it removed the start button. They put it back and things improved a bit since then but I still disklike windows 10 a lot.
8 removed it and replaced the system interface with a gesture based one even to go so far to remove it from server versions so people wouldn't put that version of explorer on their machine to get it back.
This is one of the biggest annoyances in Windows 10. If you use Command Prompt and have several of them open with overlapping windows, then it is impossible to see where one window ends and the next one starts. If you have text in all of them you'll frequently read text from the next window and then realize that text is not in the current window.
Windows 10 UX is under the control of the guy who came with the Flat UI and they don't care about annoyances such as this. They are very dogmatic, and no matter how many people complain they aren't going to give up any of their Flat UI principles.
I began using altmove after discovering xfce's alt <mousebutton> functionality. I think I had to remap the move or the resize, one of the two, in altmove to get it to behave properly.
The number of times I have opened a folder on macOS scrolled up and down wondering where the hell my file is before realizing it's hidden off to the SIDE is too damn high.
If you're talking about the "columns" view in Finder it's pretty obvious unless you're not paying attention where the contents of the folder you opened are displayed. Unless they happen to have exactly the same contents as another existing open folder at the same level the next column over's view visibly changes.
I guess this would be awkward if you've never used it, but I can't imagine it being confusing more than once or twice unless you never look at the screen while opening folders.
Icons view. Not columns. It will regularly open some of my folders in a window where the width of what it displays is less than the width of where it puts files so some require a side scroll. Why on earth it does this I have no clue.
Interesting to see that screenshot of the old IE as it was the first time we started seeing toolbar buttons that didn't have a static border to show they were a button.
You had to hover the mouse to know that it was clickable.
I agree with the rest but not this. You no longer need visible scrollbars now that you do 99% of your scrolling with the wheel or the trackpad or sometimes the keyboard. You did need visible and prominent scrollbars back when mice lacked wheels so you actually clicked and dragged them.
I still click and drag scroll bar frequently. It's faster, easier to fine control the position (especially useful when you want to toggle between two or more positions).
To me, all three ways to scroll (scrollwheel, auto-scroll, dragging scroll bar manually) have their uses.
You gotta more carefully define "vulnerable" and "computer virus" here, otherwise answers can only have equal hand-waving vagueness.
There's security related processor features that the CPUs at the time didn't have and the kind of complete sandboxing abstractions we take for granted came at a high cost (VT instruction sets were a decade off and minimum ram was 4MB, not GB, but MB ... for everything, the whole system, all of it).
As far as encryption goes, it was slow. Transparent disk encryption was totally possible but the performance hit was significant. People could be forgiven for deciding they really didn't care that much. Home users weren't keeping the kinds of secrets on their home computer that people do on their cell phones these days. The world was less digitally integrated.
Windows 95 never ran on anything less than an 80386, which could and did run reasonably secure OSes. That CPU didn't have support for hardware virtualization, but "sandboxing" is vague enough that you can reasonably say that 32-bit protected mode enforced secure sandboxing at the hardware level on those chips.
I was talking about things like Skylake's Memory Protection Extensions, Nehalems Safer Mode Extensions, Software Guard Extension, the virtualization extensions around 2005, page table virtualization with EPT, TXT, RDRAND for hardware entropy etc
That's why it depends
Things like AES and SHA are implemented in hardware these days.
We can get into discussion on whether netware or xenix on a 386 is an unusable dog or not, I've got an actual 386 about 6 feet away from me as I write this. I could even give you login if you want to live the pain.
Windows 95 made compromises, read barbarians led by Bill Gates for a good rundown. It was never intended to be their network line (NT), but instead for traditional home computers. Internet explorer was super last minute that's why it was just licensed code from spyglass mosaic. It was released the day after OEMs got windows 95.
Gates didn't think there'd be a large demand for dialup modem based internet and that they had a few years to worry about it. He pivoted effectively a month or two before release. That's why MS pushed multiple patches out quickly after the release to get a competent networking stack outside of winsock.
The people making windows 95 were competent and cared, they just had lots of conflicting priorities to attend to at once and people used it not as the manufacturers had intended. It's wasn't an easy job.
Sure it has security vulnerabilities by 2021 standards. I'd venture to say every system in 1995 that could reasonably run on a typical 4 year old home computer did at the time. 386BSD 1.0, NetBSD 1.1, Linux 1.3, likely all riddled with issues by modern standards.
And yes I'm obviously totally aware of the mainstream pop culture consensus of windows 95 sux or whatever. I mean duh, do you really think that's news to me?
I know it could BSOD with some malformed packets and that its user security was totally fake. It was totally used in ways it wasn't designed for.
Microsoft honestly thought serious stuff would be run on netware, MVS, sunos, os/2, hpux, nonstop os, tru64, os 2200, qnx, etc, certainly not windows 95. "Ah shit, here we go anyway"
Worked in an ISP whose bulk ordered install media for dialup config was preinfected either at the diskette factory or the bulk copyist. We'd shipped to customers before it was detected.
A lot of virus came from people in supply chain basically. Even extra goodies installed in the handy local pc shop, "he said it made my pc faster" still happens.
A lot of viruses were glued to the front of magazines
Extremely vulnerable. The whole time you're running the machine as root, there's simply no other option. Drivers could be in 16 or 32 bits causing instabilities and kernel-level vulnerabilities. Memory protection was easy to circumvent and no special rights are needed to listen to keyboard, record audio, video or take screenshots.
Auto-run was enabled by default. It was very easy to make bomb-cd's that installed anything just by inserting it on the drive. No centrally verified software repository and hard to acquire dev-tools forced the users to basically run whatever software was needed without compiling it themselves or by passing binaries from person to person. This alone made viruses spread vastly.
And it created bad habits that survived to this day like hiding extension of files and installation processes that encourage the user to not pay attention on what is happening and simply accept what is asked.
Windows 95 was very vulnerable to computer viruses. That being said, almost every personal computer operating system of the era was vulnerable. This was an era of single user operating systems where software could easily gain full access to the machine. The notion of the operating system running with greater privileges than applications was fuzzy since applications could simply inject a device driver override that. (Granted, that isn't technically necessary for computer viruses.)
There were, of course, exceptions (e.g. Windows NT). Those exceptions were rare in homes and small businesses. There were people who would claim that their operating system is an exception, but most of those claims were only supported by virus writers having limited interest in those platforms.
I'm convinced that the stark contrast between Bill Gates' awkward shuffle-clapping v. Steve Ballmer's cocaine-fueled mania is some kind of metaphor for the duality of man.
What manner of masochism would inspire a man to waste his time “clicking” onto icons with a peripheral whose true function is video games? — it is as though one enter text by way of a video game controller manually selecting characters one by one from a grid.
Windows 2000 is, I think, windows 95 refined, and really the peak (although some may prefer XP in classic mode).
If you tried hard, you could see bits and pieces of 3.1 interface in 95, but it was because you were running a fpga tool with an ancient file dialog, or you went out of your way to run progman, or an old installer launched progman for some reason (I think that sometimes happened, but could be misremembering 25 years ago). As opposed to newer windows where they change things, but not everything.
I think 95 was the last time Microsoft did user research to figure out how to make UI that is appealing and useful. At least, it was the last time they published anything about it. Newer stuff may be more appealing, but seems to make things require more clicks or wait to load more often, meaning it takes users longer to do tasks, and is less useful.
XP was IMO the inflection point, where Windows started to get worse faster than it got better. And close to my heart, Windows XP was the first version of Windows that was too big to fit on my trusty Compaq Presario 1210 and its whopping 1.4GB hard drive.
If Microsoft had "stopped" at Windows 2000 (like it has ostensibly "stopped" at Windows 10), and had focused on refining that piece of near-perfection, I'd probably still be a Windows user today, and happily so. My dissatisfaction with Windows - starting with XP/2003, but especially every version after - was the big reason why I looked into alternatives like Linux.
I dual-booted 2000 and XP for a time because 2000 was simply superior for meaningful work, while XP was good for gaming. And it looked like a toy, too.
The control panel was simple and clean in classic Windows, in Windows XP it's a maze with big icons and hypertext links, hopefully the classic view was still there.
Yep. Switching back to that classic view is the first thing I do on any Windows machine I touch. I've given it plenty of honest tries and in all of 5 minutes I'm fed up with it. So much easier to find things when they're, you know, actually visible and not hidden behind multiple different screens.
Discoverability in general seems to be a long-forgotten virtue of UI/UX design. Everything insists on hiding everything now. Phones do it, websites do it, macOS is increasingly doing it, Windows is increasingly doing it, Linux desktop environments are increasingly doing it (though thankfully there are a lot of holdouts)...
The "cluttered" UI is long overdue for a comeback.
> If Microsoft had "stopped" at Windows 2000 (like it has ostensibly "stopped" at Windows 10), and had focused on refining that piece of near-perfection
I for one, wished for the same thing.
Alas, once Apple came out with its shiny Aqua UI (which I must say was beautify in a totally different way), Microsoft was too insecure not to try and match that. They failed miserably in Windows XP, which felt rushed, forced, and contrived. But it certainly looked more colorful.
> Alas, once Apple came out with its shiny Aqua UI
Which is another thing that I can't stand: "Apple did something, so now we have to do it, too". Apple changed some UI element? Guess what everyone's gonna copy. Apple ditched the headphone jack? Guess what everyone's gonna copy.
You'd think that more companies would figure out that product differentiation is what makes them, you know, actually relevant amidst competition.
I can't stand it either. But I think I'm beginning to understand it.
It seems that's the way it works for competition in the consumer market. Consumers respond to the coolness factor in a major way. Microsoft probably felt that they had to try to keep up.
As near perfect as Windows 2000 was, compared to Aqua it was also perfectly ...boring.
Come to think of it, XP wasn't even a bad start given the less than two years it had in development. The real tragedy is that Microsoft couldn't do any better for the next two decades. Windows 7 was worse, 8 even worse, 10? hopeless.
Maybe it's time Microsoft throw in the towel, admit that they have no taste[0], and go back to the Windows 2000 look. They lucked into it somehow, and it still holds up better than anything else they have tried.
Things like the Office ribbon were a product of Microsoft user research, I just have no idea how they reached the conclusions they did with it (and I've watched a presentation by one of the lead designers on their reasoning - which all sounds valid, but doesn't explain how they specifically reached "the ribbon" and "personalized menus moving all over the place" as a solution).
I have a suspicion that the rise of the "user experience" engineer as opposed to UI designers played a big role - they got a bunch of people, did a bunch of research, and managed to narrow their focus to "so the first time a user ever opens Word, having somehow never used anything like it previously..." and ignored "our entire company is powered by enterprise sales of this product".
Here's my theory, having never worked there but having seen how sausages get made in a hierarchical company. A "VIP" designer in Microsoft invented the ribbon out of some creative inspiration or chemically induced fever dream. In order to justify developing it, user research had to be conducted. Since the designer was a very high ranking person, the research obviously could only have one conclusion: the ribbon is good and should be implemented. And so data was found that showed the ribbon was good, and data that showed otherwise was discarded, and like that, we got the ribbon!
Whatever the reasoning behind the ribbon, I still associate the era the ribbon was introduced with 16:9 displays becoming common, starting with laptops. In that case, users would have had less vertical screen real-estate for developers to work with. If you remember the old Office, the menus were nested fairly deeply and it became a bit of a maze to use some of the features.
I had been using computers for so long at that point that I still don't feel 100% comfortable with the ribbon. It's fine, I can do what I need to do to work the programs. It's just a bit fiddly in a way the standard menus weren't. I used to be able to invoke commands with keystrokes and arrow keys. Now there's are bunch of icons on different scales, which sometimes hide if the window isn't maximized.
Where it kicks me is the loss of real estate for toolbar buttons.
Pre-ribbon (and currently in LibreOffice which is what I now use), I could keep File control, Fonts and Styles, Reviewing and Drawing all on the one screen.
The worst is going back to mac versions of office after you have finally gotten used to everything being in the ribbon, half the stuff is in the ribbon, the other half is buried in menus, its incredibly frustrating.
The ribbon get a lot of flack but personally I love the ribbon. I fell in love with it the first time I used it, so much better than navigating through classic menus. The implementation of it is excellent in Office. Other applications have tried implementing the same thing but they don't get it as right as Office does. Many seem to think the ribbon is how you control the program, but that's not really the point of the ribbon: the ribbon is dependent on the context, it supplies you with the tools that are available for whatever part of the work piece you have in front of you.
Yes, that is one of the tenets of the ribbon. For example in Word, it makes no sense to show you options to style a table if your cursor is not inside a table.
I also felt Windows 2000 was the peak. It was still relatively simple, no "eye candy" like you had with XP, Vista, and onward.
I have an old Pentium III in the basement, running Windows 2000 server, which used to be my main desktop for a while. I haven't turned it on since 2005-ish.
IMHO Windows UI peaked with Windows 2000.[0] The main difference being that it eliminated redundant lines and reduced contrast where it wasn't needed. This improved the visual signal-to-noise ratio. I recall no significant regressions.
[0] Functionally identical to Windows XP with the luna skin disabled, as well as Windows Server 2003.
Yeah, I'm fondly remembering spending many hours just waiting for the OS to install, with me having to check back on the computer every so often just to insert disk 7 of 13. But no question it was a historical event when Win95 launched. The right culmination of factors came together in the mid-90s where now everyone needed to own a personal computer. Ink jet printers were becoming decent. Digital cameras went mainstream. CD burning with an MP3 capable stereo receiver meant you could carry 100 songs with you on the go and kids would make bank selling custom mixes for $5-$10 a pop (not everyone had a burner yet, they were expensive at the time). And of course, that whole dot com craze that turned out to be an insignificant bubble...
One of the most frustrating things for me about Windows 8/8.1/10 was the classic theme (Win95/NT style) was no longer a UI theme option without downloading something third party. It really is the thing that killed my interest in Windows after Windows 7.
Good things don't need to change. I'm thankful my remaining reason for having a Windows OS available is becoming irrelevant due to progress in running games on Linux.
I, too, always switched on Classic Theme after every Windows install. It was a large part of what drew me to XFCE/xUbuntu in the end. Numix and Adwaita-dark gets close enough to the Classic UI Windows the way I had it configured.
Recently I went through the attic and found an old laptop (120Mhz CPU, 32Mb RAM, low-RPM 810Mb HDD) with Win98 installed on in. To my amazement it could boot and start Excel and Word within 10sec. Compared to my workstation (20 cores + HT @ 2.5Ghz, 110Gb RAM, PCIe SSD) which takes upward of 40 sec to get to the desktop.
Really makes one question the current state of software. Sure, Win7 has some more features, but imagine someone in 98 saying: "we have a great idea for a new OS, it'll have more features than Win98 and it'll only take 4 times longer to boot if you are running it on a cluster of supercomputers, but those new features are gonna be worth it" =P
Sidenote on the hardware side: The laptop had a hinge for the keyboard so you could raise it to a 20 degree angle, the CD/Floppy drive had the same connector as the battery so you could swap the drive for another battery if going on a trip. There were two sliding switches for backlight and speaker-volume. Access to the internals via a simple latch. It may have been a brick but it had interesting ideas. https://en.wikipedia.org/wiki/IBM_ThinkPad_760
It used to be you could hit ctrl-c to stop waiting for dhcp. But last time I ran into that on Debian, you just had to wait for it to give up; of course that was also waiting for dhcp on a wired NIC that wasn't even plugged in.
That was quite a few years ago, I'd hope it got better, but it was one of the things that made me decide to move to FreeBSD.
Not on my machine. Windows 10 takes just a few seconds (fast boot disabled because of dual boot), whereas Linux Mint takes around 20 seconds. Maybe it's because I have many USB devices attached.
It doesn't matter but shutdown matters a lot on dual boot systems, and unfortunately Linux Mint sometimes takes minutes to restart. I believe it's related to systemd and it's very annoying (sometimes I need to switch quickly because I take Zoom calls on Windows).
It's a 1800x Amd system with 32GB Ram, main partition on SSD, /home on harddisk. Windows updates take minutes, Linux updates are usually even faster (except for upgrades which are much slower than Windows upgrades).
I'll try that analyzer. But I don't think the boot is slowed down, it's just a bit slower than Windows. Shutdown takes minutes, though.
I agree. Even when Windows has "fast startup" enabled, the time from POST to actual desktop in Linux is still better and has improved significantly, and Windows still has that "tends to go slower as you start to install stuff on it". These days I can usually reach Gnome even before my _monitors_ have had time to boot.
I used to like how KDE could load a session in the background so when you logged in it was ready: for me the post-login takes as long as boot; which makes it annoying.
Not as bad as it used to be when the audio tape failed after 20 mins, but still.
What kind of janky jalopies are you folks running? It takes my machine all of ten seconds to boot, wait for me to enter my pin, and load the desktop fully.
Yeah... I mean, I bag on Windows as much as the next Linux zealot, but my Win10 gaming machine -- still an Athlon64 -- boots from cold start to usable desktop in 5 seconds. It ONLY plays Civ V, but still.
Honest question, how do you play Civ 5 on an Athlon64? I played a couple of games on my Ryzen 2700X and the end game was unbearably slow, each turn took several minutes.
It runs pretty well on medium settings at 1080p. I have to move the world around a little sometimes to get the tiles to paint. So it’s not great, but turns only take a few seconds. Unfortunately, my $4,000 MBP with a Vega 20 runs it only a little better, so I keep playing it on the PC to save the thermal wear.
My work computer is like that. It takes a while to reconnect to all the servers and get the corporate spyware up and running.
My Windows 10 PC goes from cold/dark to ready for use in, well, 20 seconds. I keep it lean.
My 1982 TI99/4a can go from boot to running a program in about 1.5 seconds, if you can hit “2” as soon as the program selector screen shows. SOLID STATE SOFTWARE
My Commodore 64 from the early 80s booted near instantly. Maybe 1 or 2 seconds. This was a cold boot from zero--no hibernation or standby. How far we have regressed!
I didn't realize at the time how fortunate I was to have a Tandy 1000HX when I was growing up. It was one of the only PCs I'm aware of which had DOS in ROM. Of course you were stuck with DOS 2.2 or something. But startup times were beyond almost every device I've used since, and certainly miles ahead of any PC.
Acorn also had RISC OS in ROM. Booted in about two seconds. Yes it was tricky when you had to upgrade but I suspect the constraints made the ROMs better quality. Certainly was rare to come across bugs etc and the performance was miles ahead of the current software we have today.
I upgrade the OS for an Acorn A3000 around 30 years ago. It wasn't too bad. Carefully remove the old ROMs from their sockets and carefully push the new ROMs into the now empty sockets. Quickest OS upgrade I've ever done.
The IBM PS/1 had MS-DOS 4.01 in ROM. You could install newer versions on the hard drive, but I remember being bitterly disappointed on installing the 2.5MB RAM expansion, because the memory test slowed the boot process down tremendously.
Yeah, that's what I was thinking. Everything from my TI-99/4a to my Apple IIc were basically like turning on the radio. Instant. But that's mostly because they didn't really do ANYTHING at startup.
Do you also remember waiting 5 minutes for a game to load from the insanely slow 1541 floppy drive? I'll gladly take a 10 second boot time with instant application start time over instant boot time with 5 minute application start time.
I’m not sure it is necessarily the drives. My hexacore desktop spends more time in POST (or how it is called these days in uefi) than in windows boot. It takes 3s from loading the boot partition to when I can enter my password and about 20-30s before it actually decides to boot the windows EFI rom.
What annoys me isn't boot time per se, but the sluggines that permains after login as the services-not-services start up and one after another try to steal bandwidth from a WiFi that isn't yet connected.
If I could wait whatever time and then log into a usable system it'd be fine, but no, it's a minute after login before I'm in control of local resources
Hah I never noticed that those were those notes! I believe that's just a remarkable coincidence because the first, lower note is just the result of the SN76489 sound chip being powered up but "uninitialised", as it were. In theory different chips would give different tones but in practice all the ones they used seemed to give a consistent result.
> which takes upward of 40 sec to get to the desktop.
My ancient Core i5 with a whopping 4 cores gets to the Windows 10 desktop in just a few seconds.
The truth is that people care much more about features than performance -- especially system/app start up performance. Microsoft has always bet on Moore's law -- building software just not quite usable on modern hardware knowing that eventually hardware will catch up. Anyone designing software for today is immediately behind tomorrow.
weird, I got a much smaller workstation (like dual core and 10 Gb of RAM) and I get to the desktop in ~ 10 seconds. Maybe Windows is spinning up each core one after the other?
In my experience as a performance engineer, slow performance in things like boot isn't because of "bloat", it's just because of random bugs that may or may not correlate to having added more features. There is a tradeoff in development time between features and bug fixing of course.
I had that for a while on my Arch Linux install. Then I set up full disk encryption. So now of my 45s boot, 20s is me typing the password, 20s is running the luks kdf, and 5s is the entire rest of the boot process.
Counterpoint: Boot time is basically irrelevant once you have a reliable sleep/hibernate/whatever mode. My PC only reboots for windows updates or if I'm troubleshooting some inexplicable behaviour.
For the actual practical use case of sitting down at the computer and starting to use it, my PC wakes up essentially instantly. I simply don't care about boot time. My 2021 experience is far, far better than my 1998 one (well, the computer bits at least).
Mostly agree. I will concede boot time is a bit slow on 10 for me, but still faster than 95 was, and yes, it doesn't matter much because of the incredible uptime. 95 would often crash multiple times PER DAY. Meanwhile I just had my Win 10 system do an honest to goodness hard blackscreen crash a few days ago, and it was probably the first in months.
For what it's worth, I'm sure you could get the "boot" of a Win 10 machine down to almost nothing if you really wanted to. Mostly it's just not worth it.
I remember when even a cold start would drop me into a working desktop within a few seconds. Thus was in the early 90's with RISC OS running on the early days of ARM chips. Respectfully I would argue hibernation and suspend are just a band-aid for the sorry state of hardware and software today. What if we didn't need hibernation or suspend?
I had to enter my password 3 times to login, that's most of time spent here :) (8 to 10 seconds to the login banner).
I think you have an Intel machine. For some reason, Intel machines spend large amounts of time initializing their firmware. Xeon servers spend minutes at it, while Epyc ones spends only ten seconds or so. But even Intel laptops seems slower than Ryzen ones for some reason.
Edit : I had an update and rebooted (password entered once this time :)
Startup finished in 2.557s (firmware) + 799ms (loader) + 2.901s (kernel) + 5.103s (userspace) = 11.362s
graphical.target reached after 5.093s in userspace
Login banner appeared a little less than 8s after boot.
Isn't this also treating the symptom rather than providing a cure? Keeping important state only in volatile memory and hoping user clicks Ctrl+S often enough? Just resume wherever the user left off regardless if its a cold boot or not.
Androids Activity-model provides a more modern solution to this (putting its lack of multitasking aside). Did you ever think of what happens when you swap app and did you ever care about shutting off old apps? There is no need, it automatically hibernates individual apps that were not used recently but you don't notice, next time you open it magically continues exactly where you left off. Assuming developer of the app cared about state-management.
But it does play a part if you are forced through lengthy updates and reboots.
I was stuck on a 1.3mbits/s link and waited a day for a dist update that never happened.
I walked away from Windows then, and came back 6 years later. Win 10 on spinner and older hardware with 2gb ram. This originally was touted as a frugal os.
Took a day and a half of updates on a faster connection.
The machine then was so slow and CPU sucking when idle, I couldn't do anything with it. A browser would down it. Totally unusable.
Gave in, had Debian on it within an hour and it is okay with chromium, but still 2gb limits it heavily. But at least it works.
Casts mind back to win98 era, and same hardware would fly.
So if stable, and no reboots don't care so much.
They killed hibernate in Windows for a faster boot didn't they?
I have vista on a ten year laptop, and that is quite fast to boot, but now unsupported and breaks.
I know this is vaguely off topic. But main point was that boot up times aren't bothersome for stable systems with long uptime. When boot loops feel almost mandatory to get anywhere, with pretty much a virgin OS, and it takes days to actually get anywhere - it is an issue.
Actually, I liked the >1min boot times of old Linux distributions (thinking SuSE 9.0) far better than the 'you-never-know-if-you-actually-rebooted' state of modern Windows.
I think the there have been a few really important changes over the years, that make the modern computing better (like better process scheduling, that keeps your mouse smooth even when the CPU is at 100%), but the mass adoption of hibernate is not part of what I value most.
Well yes, corporate machines are a different issue.
I have a corporate mac that's sluggish due to mcafee chewing 100% CPU for 5 minute intervals, anything with "enterprise" management software on it is going to suck regardless of OS.
Just wait until you see windows 10 on a 2GB machine with mcafee, and then corona forces ms teams on it. Boot takes 15 minutes to usable. Usable means msword produces a character on the screen in less than 10 seconds after you type it.
IT closed the users ticket without action: As long as it boots in 15 minutes or less, there is no performance problem.
Yeah I have a Mac Book with 32G iof RAM and 8/16 cores and always five or so of the top 7 processes for CPU are spy ware. Unless I do a massive multi account terraform apply - that can outcompete them. Even compiling doesn’t, go is so fast.
I have to concur with the other commenters, it sounds like your system might have some kind of strange configuration. I haven't had a computer that took longer than around 10 seconds to boot since Windows 7. SSDs made a massive difference. Then Windows 8 introduced hybrid boot for certain configurations, and Windows 10 made it pretty much standard. My Android phone (Pixel) takes longer to boot than my Windows computer (Surface).
I think it's worth mentioning the reason for this is mostly because of the fast boot feature, which means Windows is actually just leaving hibernation opposed to cold booting.
Some optimized laptops are pretty insane. I just got an Asus G14 with a Ryzen 4900HS, the UEFI boot is animated and has a slicing sound and gets into desktop in maybe 5 seconds. I think I can even change the boot animation with the "ROG" button and can make the lid lights dance around unlike the lame Apple logo.
They did, but they had to turn "shutdown" into some kind of "hibernate on steroid". Doing a clean boots takes a bit longer. And we're talking modern, SSD-equipped machines vs. old laptops with spinning disks.
Shutdown, Hibernate, and Sleep are three different things. I'm talking about a proper shutdown; I almost never even bother with Hibernate because a regular cold boot is fast enough
I don't think you understand the comment parent made.
Windows 10 comes with a feature called Fast Boot and it is usually enabled by default depending on RAM, driver support etc.
What used to be a "proper shutdown" is now a hybrid hybernate. Parts of the OS in particular related to bootup are stored in a hybernation file which is loaded at boot time resulting in faster boots.
This is because Windows can now skip driver init and load state from previous boots.
I believe that cameras do that to the extreme (at least my now comically outdated G1 does that, I think): do a full reboot + hibernate on shutdown, which can take multiple seconds but which is fine because it's happening at a time when nobody is waiting for it, to then "boot up" in a fraction of a second because the system image is so small.
This is now my mental model for tricks like fast boot (which are obviously not applied to the full stack)
Honestly I've never noticed a difference on any of my Nikon DSLRs between waking up from stand-by and switching it off. (And I even had a D1 at some point!)
It's just instantly on and working. I also never noticed a difference in battery life, so I only used the on-off switch as a sort of button lock. I suspect that for Nikons that's really all it does, keep the camera from waking up from a button press.
I'd guess these use some fairly small, low power RAM that's for the "control plane", which is simply always on because power consumption is negligible, while the "data plane" is just a "dumb" ASIC with some DRAM which can turn on in a fraction of a second.
My Pana has noticeable shutdown lag (about 1.5s) suspiciously contrasting the immediate bootup/wake-up. I'd never have started suspecting that they do that hibernate trick if the shutdown started with screen-off instead of proudly (?) displaying a branded placeholder screen. Makes me wonder if maybe far more embedded systems that we know actually do have a slow boot process, but hide it in a slow shutdown that they run invisibly after screen off. If you are as curious as me, take one of your Nikon and flick the power-switch off-on quickly, that should show how deep it really went into off-ness.
If I hold the shutter down and flick the power switch on, it immediately takes a picture, if I flick back and forth it keeps taking pictures. There seems to be something like a 0.2 s delay between the click of the power switch and the clack of the shutter. That's more than the normal shutter lag (around ~50 ms for most DSLRs), but not much. If I do it two-handed I get around 2 fps out of the camera.
The Z6 mirrorless which I also have is noticeably slower. I'd say around ~1.5 seconds from the power switch to the shutter. From the noises it makes I'm guessing it is doing something with the IBIS in that extra time, which the DSLR does not have.
Power-up (remove battery, move power switch to "on" position, insert battery, observe time until meter (DSLR) or EVF (mirrorless) turns on) takes both of these around 1.5-2 seconds.
Edit: Old film Nikons also had pretty clever "power management". For example, on the Nikon FE the meter circuit is switched on and off automatically with the advance lever: If the advance lever is flush with the body, the cameras is off and the shutter is locked. Moving it to the shooting position (where it sticks about a centimeter out) the meter is powered and the shutter is unlocked. Even while advancing the film the meter is powered off. Very clever. Batteries last forever in it.
I don't know why, but I enjoy this little round of remote reverse engineering, thanks for playing along. "Cycling with shutter button pressed" is a very interesting technique, particularly when pointing at some digital stopwatch as timecode or something like that.
Another interesting specimen is the tiny RX0 I use for quick snapshots on the go, it seems to light up "services" to the screen in almost random visible order, with noticeable differences between a "doubleclick" restart and starting from settled state (e.g. screen lights up much later in the sequence on quick turnaround)
Too bad they also added Cortana, telemetry, advertising and all sorts of annoyances. It would be nice if there was a safe (no warez) way of disabling, or better removing all of them for good. Win 10 would still have an atrocious UI compared to everything since XP, but there are mods for that.
Cortana is basically dead. Microsoft has finally acknowledged her unpopularity. You can disable it easily in Windows. I went over my privacy settings with a fine-toothed comb. Haven't seen an ad or Candy Crush since the first few minutes of a new install. Uninstalled OneDrive since I don't use it and it's one of the bigger offenders for unwanted advertising.
Many, if not most, of Windows annoyances can be blocked without 3rd party software. It follows the usual Windows tradition of first look for a well hidden setting in Control Panel / Settings. If it's not there, look in Group Policy. And as a last resort, there's always a registry edit that can be done.
Telemetry is one of the annoyances which is trickier. I've seen methods which can be used without 3rd party software, but I personally just gave up since the most basic level doesn't send much data. And I use a Start Menu replacement since there's no other way of getting the old Start back.
A similar thing that really opened my eyes: running Windows XP in VirtualBox on a $999 Macbook Air from 2016.
It really flies!!! It's fast to boot up and fast to use. It's more usable than the host OS !!! It even recognizes some USB devices quite well and has drivers.
In 2021, the host OS of my Macbook Air has somehow rotted, even though I installed like 4 programs on it ever, and I barely use it (I use Linux for almost all tasks). Everything is slow, with both jank and "intended" animations.
On the other hand, virtualized Windows XP is extremely responsive.
It also runs in a tiny amount of RAM -- I think I can set the VirtualBox VM to 256 MB RAM or something (don't quote me on that)
This Macbook Air has 8 GB of RAM and it's slow. I think when I got it in 2016, it was decently fast. Somehow in 2021 all the updates made it slow to a crawl. As always iTunes is a big offender. Somehow playing MP3s has to be a chore. I think Spotlight is another offender and that's something that Windows XP didn't have.
I use Ubuntu for my Linux desktop, and it has similar problems. There's lag everywhere and it uses a ton of RAM. It's not more usable from a UI perspective. I see a lot of "lateral" progress, although to be fair Ubuntu from 10-15 years ago was pretty bad (definitely less usable than Windows), and they have made progress.
That does amaze me. I remember booting taking ages on my 166MHz Windows 98 machine. I'd press the power button and go do something else while it booted.
What used to get me was that I'd have to reboot on nearly every install/update/significant change to my PC, thus waiting hours total on just booting every month.
Not to mention just maintenance reboots because eventually the computer would just get nutty if you didn’t periodically reboot it. Some time around windows 2000 or perhaps even Vista was when you finally could run the OS for weeks without rebooting.
That looks like Windows XP based on the desktop icons and I'm quite sure that Windows 95 didn't support hibernation, it only arrived with Windows 2000 if I recall correctly.
Windows XP only came out in 2001, so if the laptop was from 1999 then it must have been upgraded to Windows XP at some point.
That text based progress bar shows exclusively for resume from hibernate, hence rubatuga saying it's "definitely hibernation".
There is NO WAY an HDD laptop boots Windows 98 in 10 seconds. Even booting Windows 98 in a VM on a 1yr old processor backed by _friggin RAM_ takes slightly more than that.
Windows 98 used to take 1-3 minutes to boot on that level of hardware. And if you had any type of network, it was probably even more than 3 minutes. There is a reason hibernation (even at the BIOS level) was all the rage these days.
For better or worse, MS did something on Windows during the XP era that finally made it be able to boot fast (on modern hardware at least). Win2K still has insanely long "delays".
Maybe has something to do with cached driver configuration, but I don't know which version of Windows introduced it.
It has to do with remembering which drivers were loaded the last time the OS ran and loading those first on reboot. Saves time on checking the whole driver catalog.
I remember at the time some propaganda about XP doing "concurrent initialization" for some drivers. My impression is they just fine-tuned some of the IO delay loops (or equivalent). Most of the Win2k boottime is spent with no CPU or disk usage. XP is way more efficient at that.
Hmm, that sounds quite a lot longer than I'm used to these days.
I just checked, and I'm 17s from cold boot to Windows 10 login prompt on my Ryzen 2600 desktop with NVMe boot drive. I also checked on one of my 5 year-old HP Zbook G3 (Xeon E3, 64GB RAM, NVMe boot drive), and it was 23s - and that's a corporate machine, loaded with crap.
> It was the time when almost everybody had a collection of MP3 files on the hard disk drive, and Winamp was one of the most downloadable Windows applications.
Objection: nobody had a collection of MP3s in 1995. The computers of 1995 didn't have the power to decompress an audio encoding like MP3 in real-time without an accelerator card. People swapped MIDI arrangements of popular songs, and some folks ripped PCM audio directly to WAV files (where it then took up half their 1GB hard disk, so this was mostly a novelty—unless you were the proud owner of a $2000 1x CD burner to burn the PCM audio back onto.)
MP3s became a big thing around 1999, with better computers and the release of Napster.
WinAmp itself was released in 1997, but wasn't immediately popular because people didn't have music on their computers to play using it. It took off around 1999 as well (which is why everyone remembers Winamp 2, rather than Winamp 1.)
Indeed. But I definitely had a collection of MP3s on my Win95 machine as soon as Napster hit the scene. I was able to fit exactly one Blink 182 album on my 1GB hard drive. By the time my mom gave me her Pentium 2 Win95 machine when she upgraded Napster had already came and went. Kazaa was the next big thing for a while, but it was never as good. CD-R/W drives were a game changer for a while though. I was amazed by the ability to load multiple CDs to make a play list and then be prompted to load the each CD into the drive to burn the tracks I wanted.
> I was able to fit exactly one Blink 182 album on my 1GB hard drive.
Dang, 1GB back in these days was pretty legit.
A 128k MP3 was something like 3-4 megs per song - I'd fit a bunch on a 100 meg Zip disk back in the day and feel like a king.
I have to disagree. A pentium at 100mhz was enough to run winamp and play 128kbps mp3 without any hiccups. Doing more than that at once was complicated, but not impossible.
It was possible, practical and happened. Collections were smaller because drives were smaller. MP3 became more popular by the end of the 90's, but a good 1995 machine, 100mhz pentium with 8mb of ram and 512mb harddisk, was 100% practical for playing mp3.
I remember there was one software decoder that was fast enough in 1995, so I plugged in my Pentium/100 laptop to decode music and pass it uncompressed to my desktop Pentium/90 (which was too slow) over 10base-T!
I used a 100mhz Pentium with 64MB or memory and, later, a 166mhz Pentium with 128MB. Win98 (people are correct that midis were way more common in the ‘95 days, as even a 2MB mp3 would take forever to download on dial-up and hard drives were small) and my experience was than any OS was fine if you were only playing mp3s, but on anything but BeOS or QNX you’d get bad pops and UI lag if you did anything else while a song way playing. Linux was actually worse about this than Windows, in fact, but both were bad.
I had a pentium mmx 200mhz with 32mb ram running windows in 1997. I could run winamp or realplayer, and at the same time I could download using get right, chat on mIRC, browse the internet using netscape and still had a file manager window open. All at the same time.
Changing from one window to another allowed me to see widgets been drawn. It wasn't snappy, but it wasn't a crawl. I could still open a dos window and run qbasic or turboC. I had to be careful: drunk pointers risked stability of the system.
At the time I never had the same thrashing problems I have today running linux. That is why I'm so hopeful about systemd-oomd.
I could keep the machine like this for almost a whole day. A windows 9x machine rarely passed 24 hours uptime with heavy usage without a bluescreen.
I later upgraded the machine to 64mb ram, 2 harddisks, 2 floopy drives and both cd drive bay became occupied: one with a cdrom drive and the other with a cd burner. I then replaced the winmodem with a network card. It was a good machine. The only thing that barred me from using linux on it was the soundcard (AZT-R2316) that had no linux driver. It 2mb s3 video card worked better with linux. By the time the sound drivers arrived it was 2007 and I already had a new linux-only 2006 machine.
With Linux, FVWM which was damn lightweight (and still is), and some "nice" sttings I was able to do it better than W98.
Rxvt was far lighter than XTerm.
Most software was for the cli after all.
Fluxbox/blackbox was even lighter. I think lightweight linux distributions, like damm small linux, could run faster than windows 9x. Unfortunately, easy to use, highly compatible, lightweight linux distros only became popular after the win9x era.
Mandrake was slow to boot on a 90's class machine compared to damn small linux. Also "run mandrake and send drakconf and KDE to the trash" doesn't ticks the "easy to use" box.
> but wasn't immediately popular because people didn't have music on their computers to play using it
I think Audiograbber was the first popular program to rip MP3's from your CD collection. Best I can tell, it was first released as v1.41 in Feb 1997. Winamp came out a couple months later.
And then young people were absolutely listening to their CD's as MP3's on their computers by the summer of 1997 -- I know I was, it was just so much more convenient. And you definitely didn't need to wait for Napster if you were on a college campus with dorms -- you'd find tons of "anonymous" network shares filled with thousands and thousands of tracks in the fall of 1997 for sure.
And the Diamond Rio player came out at the end of 1998... I remember that was huge... all of a sudden your MP3's were portable!
That's the gap between a computer enthusiast and average population. Most people weren't on a campus with a LAN. Many people were relying on 33.6K dial-up with metered a metered ISP, often on their phone line they needed to receive calls. For most people MP3s were just too hard to transfer.
Cheap CDRs were the other big breakthrough. Now people had a reason to download music so they could burn it to listen in their car. But that's well after Win95.
> between a computer enthusiast and average population
Not computer enthusiasts -- just college students.
I remember being pretty surprised, it wasn't the CS students who had all the ripped music. It was the party kids. The "killer app" for MP3's was the ability to put together all-day and all-night music mixes that could be shuffled instantly. For the first time you could DJ effortlessly.
I seem to recall a MP3 player that worked directly from DOS during that era. My computer couldn't play them in Windows, but could using the DOS based player around that time.
I'm relatively confident I had MP3 decoding software running on a 486 circa 1994/5. Even today, mpg123's website states "Plays Layer 3 in stereo on an AMD-486-120Mhz or (of course) a faster machine." and of course 48MHz ARM M4 microcontrollers can also play back MP3s.
cdda2wav seems to have existed by 1996.
the "quantum bigfoot" 1GB-and-up HDDs also in 1996 could hold several albums, especially if you ripped at lower rates like 64kBit/s.
I also recall exchanging MP3s on IRC with /dcc send, and am reasonably confident this was 1998 or earlier.
On the other hand, "On July 14, 1995, Karlheinz Brandenburg sent an e-mail announcing the .mp3 file extension had won a poll and announced that the old .bit file extension should no longer be used."
So while literally 1995 is early for people to have a "collection" MP3s on their home machines, the time before everyone had moved on _from_ Windows 95 sure overlapped with the mp3 era, especially for early adopters.
I remember our family's Mac, with a 68LC040 CPU, which was a 68040 without a dedicated floating point unit, could only play back mp2 files in realtime.
MIPS at 30Mhz is comparable to contemporary pentiums at much higher frequency of the era. The Indigo probably had faster memory, io and bigger caches than pc's mere mortals could buy. The comparison is simply not fair.
But this program called "amp"... I'm very curious to see if it could be compiled to run on a recent machine.
An Indigo sat in the office for a time playing tunes from an NFS share we would populate. I eventually took it home to do the same thing. I forget how I parted ways with that old box.
I am not inclined to build it at the moment, though it may well make for a great micro controller project one day.
I can't critically evaluate the quality since I have no high quality mp3 nor Hi-Fi audio devices. An old recording of mine http://marcodiegomesquita.tripod.com/rda.mp3 sounded with very unsaturated bass; which is good, I think I recorded/mastered it wrongly or the software I used at the time decoded it like amp decodes it. It actually sounds really good with amp.
But yes, it uses very little cpu. Maybe less than dosamp.
The reference implementation is floating-point, for accuracy, but you can bet they were using fixed-point back then, as DSPs and such would be fixed-point only.
I recall a friend having a pentium 1 at 100mhz and could only play 64-128 kbps mp3's if he did literally nothing else. The song would skip from time to time and would immediately start skipping if you moved the mouse. From what I remember this was with winamp.
My first machine was a 100mhz Pentium; Winamp used about 25% CPU to play MP3s - I did it all the time while coding and hanging out on IRC. Can't think why your friend's machine would struggle with the same setup. Maybe older versions of Winamp were underoptimized.
I used Sonique for a while because it looked amazing - slightly poorer performance, but felt worth it. Later on, Sonique "optimized" for later Pentium releases and it became much too slow on the P100, so I switched back to Winamp.
I recall 1997 being when every IRC-type person had mp3s and a cd burner. One was lucky to have a hard drive large enough to convert mp3s to wav files before burning. Otherwise you had to decompress "on the fly" and hope for the best. Could be done at 0.5x for higher chance of success, if I recall correctly.
I remember when the first burners came out that were capable of automatically resyncing and restarting after an error. I've forgotten the brand now, but there was one company that shipped it first, and instantly everyone in the LAN party scene bought it. Ricoh maybe? Burning anything was such a pain before that. Even if you took every precaution in the world, you couldn't stop a big truck from driving by and turning your burn into a coaster.
I entered college in 1995. I downloaded my first MP3 in summer of 1996 (Savage Garden -- Chicken Cherry Cola). By the end of 1997 (which is when the OP is talking about, when Winamp dropped) I had hundreds of MP3.
Scour was released in November 1997, and it was released to solve the problem of a whole bunch of college students having their MP3 libraries available via Windows shares, but it was hard to find what you were looking for. Scour fixed that.
So maybe not the general public, but in 1997 someone released a search engine specific to finding MP3s, because people already had large collections.
> So maybe not the general public, but in 1997 someone released a search engine specific to finding MP3s, because people already had large collections.
I was the author of MP3 Fiend around this time.
It was a desktop metasearch engine for MP3s that worked in conjunction with multiple download clients.
It was very successful at the time, only for Napster to come along and eat my lunch.
Ah yes, the days of mis-titled MP3 files, incorrect metadata, and other artifacts. Sometimes an MP3 of a song you liked had a click in a certain spot, or occasionally some kind of sound like an AIM message coming through.
I ran my company's (unofficial) MP3 server in '99, and I spent a LOT of time managing files.
I ultimately went back to college in 2002 which was the right time to catch Apple's first attempt at Home Sharing in iTunes (I forget what it was called), where basically every single Mac on the campus network was running an HTTP server advertising all of its music for download.
Thank you sir. My windows 95 machine could not play an MP3, and I could hardly download one in my dialup connection. My entire hard drive was 500mb, and I couldn’t have saved many MP3s even if I’d wanted. I did have about 1000 MIDI files, though.
I have a hunch that you're mixing up bits and bytes in that memory, because I very distinctly remember seeing the internet explorer 4 download progress indicator telling me that I was getting 4.5 "KBPS" down on a 48K link, which makes sense when you convert kilobits per second to kilobytes per second. Internet Explorer would give download speeds in kilobytes per second. If you were getting 3 kilobytes per second, then that's equivalent to ~28.8 kilobits per second, which my buddy who lived out in the middle of nowhere would get. He would come to my house in town so he could get better performance in Runescape. This was in 2003. I am not joking.
You have the right idea -- I wasn't mixing them up, but as you note, I expected to be able to get more than 3KBPS on a 56.6k link. I still managed to download about 3,000 songs on that link, though! I was very dedicated. I couldn't actually browse the internet while downloading, so I had to queue up webpages, or just find something else to do while I downloaded a single song.
Yes they did. Straight otta Fraunhofer! We used Winplay3 and mpg123 to get mp3playback. Winamp was a godsend when it came, due to many quality of life improvements.
I do remember using my 1997-era computer (Pentium II, 233 MHz, 32 MB RAM, 3 GB disk) to listen to audio even though I didn't have a collection of MP3s. Many computers at that time shipped with a CD-ROM (if not a CD-R or CD-RW) drive, and so could play back audio off of CDs.
I also remember waiting a few minutes to download Winamp, ICQ, and the like. (A dial-up modem would get 4 KB/s down on a good day, so 146 KB was not insignificant for an application.)
I graduated from high school in 1996, and everyone in college dorms were exchanging music files with point to point sharing in the dorms. It was early adopter era, but the computers could handle it.
Ripping CDs and encoding was definitely a different story.
> People swapped MIDI arrangements of popular songs
I recall downloading MOD files back in the day from various BBSs. This is probably pre-95 by a few years. MIDI files were good but you needed a good sound card to play them. I recall drooling over ads of the Gravis Ultrasound (I had a SB Pro).
Someone always pulls out a C64 demo. This is using a bunch of tricks to do it and it is massively compressed.
It generally isn't doable. My Amiga 600 without the Vampire or a Prisma Megamix cannot play MP3s of decent quality and the Amiga was far more capable than C64.
You are mistaken. It took a lot of cpu but definitely worked. I can't remember the name of it, but I used some quirky mp3 player for a couple years before winamp came out.
I do remember being impressed how mod / tracker based music was so much more efficient on our P5 120Mhz. It was something like 10% of cpu vs 50% for mp3.
I worked the Egghead Software midnight release party. Yes, people lined up at midnight to get a physical, boxed copy of Windows 95. Queue the Rolling Stones "start me up" that cost BillG $3 million.
there were Windows 98 midnight launches as well. People lining up outside of CompUSA to buy their copies.
3 of us had our Apple t-shirts (from the Apple "Demo Days" program) and lined up at a Sacramento, CA location (I forgot what chain it was) that had an Apple "store in a store" - media (radio) was outside and interviewed us on air about it. We fired up all the Macs, and our friend crashed the big screen demo with the c:\con\con bug.
I distinctly remember an interview on Win95 launch day where the local news was interviewing a guy in a computer shop about his thoughts. He was unimpressed. He said something like, "Mac* already does this and much better".
When the iPhone launched, my reaction to it was essentially the same as the guy in the computer shop. Boy were we wrong about the impact both would have on the industry!
> (Boot to DOS) This feature is not available more, but in Linux, the possibility to boot in a console mode still exists.
That's not really how it works - there's no "console mode" in Linux. On Linux the Desktop environment is built on top of underlying OS interface - that's precisely why you can run it completely headless, or why you can completely switch desktop environments in a few seconds.
Win95 was kind of the intermediate between win 3.1 where it really worked that way, and later windows where its definitely not on top of a cosole based os.
If you boot into dos mode on win95 its just stopping the boot process at the start windows step, which is not that different from booting linux but not starting x windows.
No, that‘s not exactly what happens, albeit a commen misconception. Win 95 (and earlier releases) were not just a GUI shell on top of DOS. They were Operating Systems (with their own functions and capabilities) that used DOS as part of their boot process and to provide compatibility for some things (oder drivers, etc.)
If you executed a DOS program within Win 95/3.1 it did not run on the underlying DOS “layer” but in some sort of DOS virtual machine.
So what happened when you booted into “dos mode”? You really booted into DOS like you said without the last step of loading Windows. But loading Windows was not just like startx but more a continuation of the boot process into the Windows Operating system.
The Linux system console is implemented directly in the kernel and is not really "on top" of anything, so I'd say that Linux having a "console mode" is a rather on-point description. This isn't always a good thing, since it includes a whole terminal-standards-emulation component that's one of the kludgiest in Linux itself.
> In 1996 the ICQ — the first so-called “instant messenger”, was released. Now it is standard to be always online and to have different chats in Slack or WhatsApp, but in 1996 it was a sort of new idea.
What? IRC was well alive back then and you could have multiple chats in a single IRC client. This claim makes no sense. What ICQ brought was not the idea of multiple chats going on, but a single identifier to connect to people you knew (and that NOT a mobile phone number)
It's about the popularity of the product. Yes, IRC is much older, but it was only for geeks. ICQ and the successors (including recent ones, such as WhatsApp) are for the masses.
IRC had zillions of web interfaces everywhere, and it
was used by everyone in Europe, at least in order
to chat in public channels based either by theme
or by regional locations.
IRC was too hard for most people to use. Also, "IRC" is not a single entity. Which network are you connecting to? What's your nick? Is it reserved? Are you running a bouncer, or can I only message you when you're online? Way too complicated.
ICQ, and later AIM, made most of this stuff "easy" for the average person. Too bad AIM didn't evolve more... It could've been Slack 20 years earlier.
The biggest surprise for me in reading this was the complete lack of any SSL functionality in the original version of Internet Explorer, rendering most of the modern web not merely broken, but fully inaccessible.
I wonder what the earliest version of Windows would be that can at least establish a connection to a 2020s website out of the box.
Out of the box you're looking at probably Vista SP1, or maybe even 7. SNI support is basically mandatory now. You could install older versions of firefox or IE8 for XP, I guess.
Windows 7 RTM cannot install the chocolatey package manager OOTB due to TLS 1.2 expectations, necessitating some work to get Windows up to date with Powershell V3 and such[1]. So maybe Windows 7 SP1, or Windows 8?
The past couple years SSL has been disappearing completely, and TLS is now required. It didn't exist until at least 1999. So even with SSL support, you'd still be out of luck.
About five or six years ago I found an old CD of Windows ME. Having moved on from Windows 95 to NT, 2000 and so forth, I was curious about it. Needless to say, even a few years ago getting on the web with Windows ME out of the box was broken. I eventually found a version of Opera I think which was new enough that I was able to use the Internet. More recently I read of someone using a SOCKS proxy to make using the Internet more functional for an older machine.
Internet Explorer v1 was not really a product. I think the only place I saw it was on a documentation CD which contained HTML files. The original internet pack for Windows 95 came with IE2, which had SSL.
> But this connection is practically useless — web standards went so far ahead during the last 25 years, 99% of websites just cannot be opened. I can ping the website, and that is mostly the maximum I can do:
You can still open http only websites fine, even in the original internet explorer : https://whynohttps.com/
"It’s a known issue that occurs on processors with a frequency higher than 2.1 GHz......in 1995 nobody was thinking that Windows will be running on a CPU with a so high frequency."
The good old days. Earlier no body would ever need more than 640K main memory....
> It was also interesting to see the beginning of the new era, the era of connected devices and online services, and to see what will happen with all these services after 10–20 years. This is something that is interesting to think about — will we be able to show our grandchildren how did old stuff work, or “Cannot connect to server” will be the only message that will be displayed?
You can't compare the very beginning of the internet and what will come from where we are now in the next 10-20 years. Most likely a lot of things will still work in 10 years, because the underlying foundations of the protocols are now standardized and fixed for a while. Same thing with cars from 30 years ago: they can run just fine on the road and with regular gas, because its matured technology.
The very beginning of the internet was the 60s not the 90s.
And honestly where protocols have changed its been very recent. There was a big lull between the 90s and last couple years. TLS 1.3, the most obvious example of something new, was very recent (1.0-1.2 were minor but important improvements to SSL 3, which until recently was still widely supported. TLS1.3 was the biggest change in a long time). Http/2 (and 3!) Are super new, and earlier versions of http which date to the 90s are still universally supported.
He probably meant web and not Internet. The IP protocol, which is arguably what is closest to what internet really means haven’t seen a change since the 80s. We are still struggling with moving to IPv6.
But his point was that the foundations were fixed. You could say the same thing about the internet in the early 90s, and then http changed a bunch of things. Whose to say the same thing won't happen tommorow. It doesn't seem like HTTP is any more of a fixed foundation today than tcp/ip was in the 90s
TCP/IP is still strong with us today. The whole web is built on top of that. Given that browsers today already support HTTP/2, it is hard to imagine that in 25 years the web would have moved to something else. HTTP/1 has been around for longer than that and it's still strong. The only change that made things incompatible was SSL, but it is unlikely that will change anytime soon either.
I think the take away from that blog post is that the user experience from 95 is not that drastically different from the experience from today. The leap from the 80s UI to 90s was drastic, but it has been getting more stable since. To the point that I would call it stagnated for at least the past 10 years. The web UI did change a lot and, as noticed, it is completely incompatible with browsers from 95. But that's largely because it was catching up with the native apps/ui. I think we are there already, though. So I expect the web to be as stable as the native protocols and graphical toolkits have been for the past 25 years. Probably even more so, since keeping backwards compatibility is much more important in the web.
Windows XP can be considered a mature operating system and is barely usable online today. It could be that a new technology renders some major components and standards obsolete. It’s simply not comparable to mechanical engineering.
It’s even more possible today, when a small number of large companies make up for a huge amount of functionality, that we’ll suddenly lose it. GameSpy Arcade was one example of how a single market shift can end entire communities. I fear IoT will be even worse.
>Windows XP can be considered a mature operating system and is barely usable online today
XP is barely usable online today, but XP itself was released (2001) before we had anything resembling the modern web. It'd be years before the phrase "Web 2.0" even entered popular lexicon.
Also there was an application, I believe called "gooey" that would automatically create a chatroom for any site you visit. So when you visit a site, there would be a chat for that particular url and people would chat in it.
It looks better and more functional than my Gtk3 based desktop environment on linux today. It's amazing how much the desktop environments have regressed in linux over the last decade.
I have used something other than GNOME for more than a decade. The GNOME devs still control Gtk3 and the Gtk3 file chooser is used by many DE and applications. Heck, I can't even get my KDE/Qt applications to use their native file chooser under Debian 10 w/MATE. So even using KDE applications is no escape.
> Is 100 MB is the minimum program size in 2021? Last time, when I was downloading the drivers for my new Brother printer, it was about 250 MB, I still have no idea what they did include in the archive.
When I see something that should be tiny being about 250MB, it's usually that they bundled Qt for the GUI. I suppose that's a cost of cross-platform tools. The 10kb executable the author made was using system GUI libraries, a dynamically linked executable on a system with Qt etc already installed would be similar.
They've probably only included the modules they're using, or maybe they're statically linked and so only including the specific code used. Bundling the whole set of shared libraries is what I've seen a lot of though! Often in the context of a Python program using PyQt.
524 comments
[ 3.1 ms ] story [ 326 ms ] threadAdding things like transparency or rounded corners breaks a lot of that — you need to know what's underneath to update the display — and that wasn't fast enough to use until you had enough RAM to store each object independently so you don't need to ask each application to repaint as things move around. Compositing window managers started early — I believe the Amiga had that — but the first mainstream OS was when Mac OS X shipped Quartz and it had performance drawbacks (i.e. tolerable but not great window dragging performance) until they shipped Quartz Extreme which used OpenGL to do the hard work. Microsoft caught up about 5 years later with Windows Vista, in part because they had more users to support with inadequate hardware.
I agree that roundness wasn’t prohibitive in general. The rectangular nature of the APIs encouraged it but you certainly saw it in applications. What I was thinking about most was complex usage where there was overlap outside of the active context (e.g. dragging a window with the expectation of a lower layer showing around the curve, rather than either not having that or being able to assume a solid fill color) or interfering with hardware accelerated painting or later video decoding which only worked in rectangular regions.
* For it's time. I'm sure it wasn't instant on a pentium 1 even if I remember it that way.
OpenBSD has canfield (solitaire), cribbage, [go]fish, mille [bornes], and even monop[oly] and bcd (format input as punch cards) out of the box (and way more! this is just a fraction of intro(6): http://man.openbsd.org/man6/intro.6). Unfortunately I’m not aware of any GNU distros that do, but you can get them with a simple `sudo apt install bsdgames` on Debian (also emacs has tons of games of course https://www.emacswiki.org/emacs/CategoryGames).
Slackware does: https://mirrors.slackware.com/slackware/slackware64-current/...
Even Debian, if you're not careful, can become a mess. Windows is dumb enough that you don't need to be careful.
Chocolatey was the last thing I missed from Linux/Unix distros.
And I can still code my own card game if I really want to :)
EDIT: seeing a sibling talking about slackware, it seems I missed the joke ^^
As it is now, and forever will be, I think the balance between usability and ease of development will be those 20% of effort to make your app palatable to 80% of the users (barring some exceptions by people with underdeveloped moneymaking sense whose self-description might involve the word "artisanal".)
- Menus
- Easy keyboard shortcuts with underlined keys in menus (Alt-F Alt-O!), and reconfigurable in general,
- F1 Help working everywhere,
- Unified keyboard shortcuts: F2 to rename, F5 to refresh (except in Lotus which closes entirely),
- Reconfigurable an draggable toolbars!
- Icons (those don’t exist in the web)
And so on. Treeviews! Lists with sortable and reorderable headers! They were awesome. It was the ultimate UI. Then they flattened everything and, worse than everything, came the time where treeviews disappeared and were replaced with infinite list sorted by “last document opened” like in Google Drive... That is what I regret most: Old documents being buried behind an infinite scroll, you can never be sure that you see and sort everything.
Remember tables with sortable column headers, in every app?
Oh man. The web sucks so hard, and the kids working on it now are so used to it they don't even know how hard they've been fucked over. To think that in 2021 people still have to think about how to wire up a data table to some backend, thousands of people writing their own version of that every year....
Imagine. The computer giving you a list of the actions you can do, written in a language you already understand, organized hierarchically by the kind of thing you are operating on. And as a bonus, most of them have keyboard shortcuts that are/were immediately spelled out directly on the menu!
The other day, I was using the Twitter web interface, and I realised I'd accidentally clicked the "Like" button on a tweet as I was going back to the previous page. Oh no, I haven't liked a bad tweet, have I? Of course, I go forward, and the page has completely changed, showing me a bunch of stuff that wasn't there the last time I saw it. Great, now I need to navigate the UI. I scan the interface for some sort of "likes list". It's not in the sidebar. It's not in the "more" in the sidebar. It's not even in "lists".
So I open Tweetbot, search the menu bar for "like", find it in "Tabs → Likes", and perform the action I want in seconds. Web applications can never compare.
No it does NOT. It's beautify because it's simple, yet consistent and clear.
Today's "standard"? That's the ugly one.
Win 10 brought their current start menu.
These days, dark mode on a black desktop background means you can't find where a window's surface ends.
Windows 10 UX is under the control of the guy who came with the Flat UI and they don't care about annoyances such as this. They are very dogmatic, and no matter how many people complain they aren't going to give up any of their Flat UI principles.
https://edition.cnn.com/2021/01/05/tech/windows-10-redesign-...
I began using altmove after discovering xfce's alt <mousebutton> functionality. I think I had to remap the move or the resize, one of the two, in altmove to get it to behave properly.
http://www.altmove.com/
I guess this would be awkward if you've never used it, but I can't imagine it being confusing more than once or twice unless you never look at the screen while opening folders.
You had to hover the mouse to know that it was clickable.
I agree with the rest but not this. You no longer need visible scrollbars now that you do 99% of your scrolling with the wheel or the trackpad or sometimes the keyboard. You did need visible and prominent scrollbars back when mice lacked wheels so you actually clicked and dragged them.
To me, all three ways to scroll (scrollwheel, auto-scroll, dragging scroll bar manually) have their uses.
There's security related processor features that the CPUs at the time didn't have and the kind of complete sandboxing abstractions we take for granted came at a high cost (VT instruction sets were a decade off and minimum ram was 4MB, not GB, but MB ... for everything, the whole system, all of it).
As far as encryption goes, it was slow. Transparent disk encryption was totally possible but the performance hit was significant. People could be forgiven for deciding they really didn't care that much. Home users weren't keeping the kinds of secrets on their home computer that people do on their cell phones these days. The world was less digitally integrated.
That's why it depends
Things like AES and SHA are implemented in hardware these days.
We can get into discussion on whether netware or xenix on a 386 is an unusable dog or not, I've got an actual 386 about 6 feet away from me as I write this. I could even give you login if you want to live the pain.
Windows 95 made compromises, read barbarians led by Bill Gates for a good rundown. It was never intended to be their network line (NT), but instead for traditional home computers. Internet explorer was super last minute that's why it was just licensed code from spyglass mosaic. It was released the day after OEMs got windows 95.
Gates didn't think there'd be a large demand for dialup modem based internet and that they had a few years to worry about it. He pivoted effectively a month or two before release. That's why MS pushed multiple patches out quickly after the release to get a competent networking stack outside of winsock.
The people making windows 95 were competent and cared, they just had lots of conflicting priorities to attend to at once and people used it not as the manufacturers had intended. It's wasn't an easy job.
Sure it has security vulnerabilities by 2021 standards. I'd venture to say every system in 1995 that could reasonably run on a typical 4 year old home computer did at the time. 386BSD 1.0, NetBSD 1.1, Linux 1.3, likely all riddled with issues by modern standards.
And yes I'm obviously totally aware of the mainstream pop culture consensus of windows 95 sux or whatever. I mean duh, do you really think that's news to me?
I know it could BSOD with some malformed packets and that its user security was totally fake. It was totally used in ways it wasn't designed for.
Microsoft honestly thought serious stuff would be run on netware, MVS, sunos, os/2, hpux, nonstop os, tru64, os 2200, qnx, etc, certainly not windows 95. "Ah shit, here we go anyway"
The curse of success
Isn't this the hallmark of all truly revolutionary technologies?
Malware and AUTORUN.EXE: name a more iconic duo.
Many Win 95 computers had only sporadic net access or none at all.
A lot of virus came from people in supply chain basically. Even extra goodies installed in the handy local pc shop, "he said it made my pc faster" still happens.
A lot of viruses were glued to the front of magazines
Auto-run was enabled by default. It was very easy to make bomb-cd's that installed anything just by inserting it on the drive. No centrally verified software repository and hard to acquire dev-tools forced the users to basically run whatever software was needed without compiling it themselves or by passing binaries from person to person. This alone made viruses spread vastly.
And it created bad habits that survived to this day like hiding extension of files and installation processes that encourage the user to not pay attention on what is happening and simply accept what is asked.
There were, of course, exceptions (e.g. Windows NT). Those exceptions were rare in homes and small businesses. There were people who would claim that their operating system is an exception, but most of those claims were only supported by virus writers having limited interest in those platforms.
What manner of masochism would inspire a man to waste his time “clicking” onto icons with a peripheral whose true function is video games? — it is as though one enter text by way of a video game controller manually selecting characters one by one from a grid.
If you tried hard, you could see bits and pieces of 3.1 interface in 95, but it was because you were running a fpga tool with an ancient file dialog, or you went out of your way to run progman, or an old installer launched progman for some reason (I think that sometimes happened, but could be misremembering 25 years ago). As opposed to newer windows where they change things, but not everything.
I think 95 was the last time Microsoft did user research to figure out how to make UI that is appealing and useful. At least, it was the last time they published anything about it. Newer stuff may be more appealing, but seems to make things require more clicks or wait to load more often, meaning it takes users longer to do tasks, and is less useful.
XP was IMO the inflection point, where Windows started to get worse faster than it got better. And close to my heart, Windows XP was the first version of Windows that was too big to fit on my trusty Compaq Presario 1210 and its whopping 1.4GB hard drive.
If Microsoft had "stopped" at Windows 2000 (like it has ostensibly "stopped" at Windows 10), and had focused on refining that piece of near-perfection, I'd probably still be a Windows user today, and happily so. My dissatisfaction with Windows - starting with XP/2003, but especially every version after - was the big reason why I looked into alternatives like Linux.
Discoverability in general seems to be a long-forgotten virtue of UI/UX design. Everything insists on hiding everything now. Phones do it, websites do it, macOS is increasingly doing it, Windows is increasingly doing it, Linux desktop environments are increasingly doing it (though thankfully there are a lot of holdouts)...
The "cluttered" UI is long overdue for a comeback.
I for one, wished for the same thing.
Alas, once Apple came out with its shiny Aqua UI (which I must say was beautify in a totally different way), Microsoft was too insecure not to try and match that. They failed miserably in Windows XP, which felt rushed, forced, and contrived. But it certainly looked more colorful.
Which is another thing that I can't stand: "Apple did something, so now we have to do it, too". Apple changed some UI element? Guess what everyone's gonna copy. Apple ditched the headphone jack? Guess what everyone's gonna copy.
You'd think that more companies would figure out that product differentiation is what makes them, you know, actually relevant amidst competition.
It seems that's the way it works for competition in the consumer market. Consumers respond to the coolness factor in a major way. Microsoft probably felt that they had to try to keep up.
As near perfect as Windows 2000 was, compared to Aqua it was also perfectly ...boring.
Come to think of it, XP wasn't even a bad start given the less than two years it had in development. The real tragedy is that Microsoft couldn't do any better for the next two decades. Windows 7 was worse, 8 even worse, 10? hopeless.
Maybe it's time Microsoft throw in the towel, admit that they have no taste[0], and go back to the Windows 2000 look. They lucked into it somehow, and it still holds up better than anything else they have tried.
One can always hope.
[0] https://www.cnet.com/news/steve-jobs-our-favourite-quotes/
I have a suspicion that the rise of the "user experience" engineer as opposed to UI designers played a big role - they got a bunch of people, did a bunch of research, and managed to narrow their focus to "so the first time a user ever opens Word, having somehow never used anything like it previously..." and ignored "our entire company is powered by enterprise sales of this product".
I had been using computers for so long at that point that I still don't feel 100% comfortable with the ribbon. It's fine, I can do what I need to do to work the programs. It's just a bit fiddly in a way the standard menus weren't. I used to be able to invoke commands with keystrokes and arrow keys. Now there's are bunch of icons on different scales, which sometimes hide if the window isn't maximized.
Pre-ribbon (and currently in LibreOffice which is what I now use), I could keep File control, Fonts and Styles, Reviewing and Drawing all on the one screen.
Post-ribbon...this is literally impossible.
Is it? That would explain why I find it confusing and moving buttons around randomly.
I have an old Pentium III in the basement, running Windows 2000 server, which used to be my main desktop for a while. I haven't turned it on since 2005-ish.
[0] Functionally identical to Windows XP with the luna skin disabled, as well as Windows Server 2003.
But it also came with a music video from Weezer which played FULL SCREEN. And felt like full on witchcraft after the pixelly goodness of windows 3.11
Superficial churn like flat, skeumorphic, minimalist, whatever, sure, but nothing like TUI to GUI, windows, or the mouse.
Good things don't need to change. I'm thankful my remaining reason for having a Windows OS available is becoming irrelevant due to progress in running games on Linux.
Which bit of 95 did you fail to understand?
That part.
Really makes one question the current state of software. Sure, Win7 has some more features, but imagine someone in 98 saying: "we have a great idea for a new OS, it'll have more features than Win98 and it'll only take 4 times longer to boot if you are running it on a cluster of supercomputers, but those new features are gonna be worth it" =P
Sidenote on the hardware side: The laptop had a hinge for the keyboard so you could raise it to a 20 degree angle, the CD/Floppy drive had the same connector as the battery so you could swap the drive for another battery if going on a trip. There were two sliding switches for backlight and speaker-volume. Access to the internals via a simple latch. It may have been a brick but it had interesting ideas. https://en.wikipedia.org/wiki/IBM_ThinkPad_760
Fast forward to now and I'll hit the power button, go do some stuff for a few minutes, and then finally use the damn thing.
That was quite a few years ago, I'd hope it got better, but it was one of the things that made me decide to move to FreeBSD.
It doesn't matter but shutdown matters a lot on dual boot systems, and unfortunately Linux Mint sometimes takes minutes to restart. I believe it's related to systemd and it's very annoying (sometimes I need to switch quickly because I take Zoom calls on Windows).
Windows updates still take hours and still sometimes need multiple reboots vs Kubuntu which updates whilst in use and barely ever needs rebooting.
Having come back to MS Win (for work) after a decade+ away, and with an MS produced computer I was expecting to be blown away by the speed.
Aside: have you tried systemd-analyze blamw, very useful for diagnosing boot slowdowns; https://www.commandlinux.com/man-page/man1/systemd-analyze.1...
I'll try that analyzer. But I don't think the boot is slowed down, it's just a bit slower than Windows. Shutdown takes minutes, though.
Not as bad as it used to be when the audio tape failed after 20 mins, but still.
I have two laptops. One I own, with Windows 10, and one from work, also with Windows 10. Both Thinkpads, but not the same model.
I own the cheaper one. It’s faster.
My Windows 10 PC goes from cold/dark to ready for use in, well, 20 seconds. I keep it lean.
My 1982 TI99/4a can go from boot to running a program in about 1.5 seconds, if you can hit “2” as soon as the program selector screen shows. SOLID STATE SOFTWARE
Here's a quick demo from a similar Tandy model: https://www.youtube.com/watch?v=DaSkda4XW3k
It's in ARM assembler.
I have two PCs, one with a 2.5" SSD and another with an m.2, both are ready to use with 10-15 seconds of cold boot.
But cold boot usually isn't necessary, since they sleep just fine.
If I could wait whatever time and then log into a usable system it'd be fine, but no, it's a minute after login before I'm in control of local resources
But then again it had a lot less in terms of hardware to deal with, memory to test and so on.
To be fair to PCs the DFS prompt was more like a boot-loader than an OS.
My ancient Core i5 with a whopping 4 cores gets to the Windows 10 desktop in just a few seconds.
The truth is that people care much more about features than performance -- especially system/app start up performance. Microsoft has always bet on Moore's law -- building software just not quite usable on modern hardware knowing that eventually hardware will catch up. Anyone designing software for today is immediately behind tomorrow.
For the actual practical use case of sitting down at the computer and starting to use it, my PC wakes up essentially instantly. I simply don't care about boot time. My 2021 experience is far, far better than my 1998 one (well, the computer bits at least).
For what it's worth, I'm sure you could get the "boot" of a Win 10 machine down to almost nothing if you really wanted to. Mostly it's just not worth it.
When I look at my PC it is so painful to see, that it actually spends more time with EFI than with the OS:
Why can't they just open-source whatever they crapped out, and have professional developers maintain it for them for free?
I had to enter my password 3 times to login, that's most of time spent here :) (8 to 10 seconds to the login banner).
I think you have an Intel machine. For some reason, Intel machines spend large amounts of time initializing their firmware. Xeon servers spend minutes at it, while Epyc ones spends only ten seconds or so. But even Intel laptops seems slower than Ryzen ones for some reason.
Edit : I had an update and rebooted (password entered once this time :)
Startup finished in 2.557s (firmware) + 799ms (loader) + 2.901s (kernel) + 5.103s (userspace) = 11.362s graphical.target reached after 5.093s in userspace
Login banner appeared a little less than 8s after boot.
Androids Activity-model provides a more modern solution to this (putting its lack of multitasking aside). Did you ever think of what happens when you swap app and did you ever care about shutting off old apps? There is no need, it automatically hibernates individual apps that were not used recently but you don't notice, next time you open it magically continues exactly where you left off. Assuming developer of the app cared about state-management.
I was stuck on a 1.3mbits/s link and waited a day for a dist update that never happened.
I walked away from Windows then, and came back 6 years later. Win 10 on spinner and older hardware with 2gb ram. This originally was touted as a frugal os.
Took a day and a half of updates on a faster connection.
The machine then was so slow and CPU sucking when idle, I couldn't do anything with it. A browser would down it. Totally unusable.
Gave in, had Debian on it within an hour and it is okay with chromium, but still 2gb limits it heavily. But at least it works.
Casts mind back to win98 era, and same hardware would fly.
So if stable, and no reboots don't care so much.
They killed hibernate in Windows for a faster boot didn't they?
I have vista on a ten year laptop, and that is quite fast to boot, but now unsupported and breaks.
I think the there have been a few really important changes over the years, that make the modern computing better (like better process scheduling, that keeps your mouse smooth even when the CPU is at 100%), but the mass adoption of hibernate is not part of what I value most.
I suspect configuration (software and hardware) has a massive impact which is being glossed over.
I have a corporate mac that's sluggish due to mcafee chewing 100% CPU for 5 minute intervals, anything with "enterprise" management software on it is going to suck regardless of OS.
IT closed the users ticket without action: As long as it boots in 15 minutes or less, there is no performance problem.
You should try Windows 10. They made dramatic improvements to the boot time and my (significantly less beefy) desktop cold-boots in <10s.
This is now my mental model for tricks like fast boot (which are obviously not applied to the full stack)
It's just instantly on and working. I also never noticed a difference in battery life, so I only used the on-off switch as a sort of button lock. I suspect that for Nikons that's really all it does, keep the camera from waking up from a button press.
I'd guess these use some fairly small, low power RAM that's for the "control plane", which is simply always on because power consumption is negligible, while the "data plane" is just a "dumb" ASIC with some DRAM which can turn on in a fraction of a second.
The Z6 mirrorless which I also have is noticeably slower. I'd say around ~1.5 seconds from the power switch to the shutter. From the noises it makes I'm guessing it is doing something with the IBIS in that extra time, which the DSLR does not have.
Power-up (remove battery, move power switch to "on" position, insert battery, observe time until meter (DSLR) or EVF (mirrorless) turns on) takes both of these around 1.5-2 seconds.
Edit: Old film Nikons also had pretty clever "power management". For example, on the Nikon FE the meter circuit is switched on and off automatically with the advance lever: If the advance lever is flush with the body, the cameras is off and the shutter is locked. Moving it to the shooting position (where it sticks about a centimeter out) the meter is powered and the shutter is unlocked. Even while advancing the film the meter is powered off. Very clever. Batteries last forever in it.
Another interesting specimen is the tiny RX0 I use for quick snapshots on the go, it seems to light up "services" to the screen in almost random visible order, with noticeable differences between a "doubleclick" restart and starting from settled state (e.g. screen lights up much later in the sequence on quick turnaround)
Many, if not most, of Windows annoyances can be blocked without 3rd party software. It follows the usual Windows tradition of first look for a well hidden setting in Control Panel / Settings. If it's not there, look in Group Policy. And as a last resort, there's always a registry edit that can be done.
Telemetry is one of the annoyances which is trickier. I've seen methods which can be used without 3rd party software, but I personally just gave up since the most basic level doesn't send much data. And I use a Start Menu replacement since there's no other way of getting the old Start back.
A sibling comment already covers how you can remove/disable nearly every preinstalled annoyance so I won't go into that
It really flies!!! It's fast to boot up and fast to use. It's more usable than the host OS !!! It even recognizes some USB devices quite well and has drivers.
In 2021, the host OS of my Macbook Air has somehow rotted, even though I installed like 4 programs on it ever, and I barely use it (I use Linux for almost all tasks). Everything is slow, with both jank and "intended" animations.
On the other hand, virtualized Windows XP is extremely responsive.
It also runs in a tiny amount of RAM -- I think I can set the VirtualBox VM to 256 MB RAM or something (don't quote me on that)
This Macbook Air has 8 GB of RAM and it's slow. I think when I got it in 2016, it was decently fast. Somehow in 2021 all the updates made it slow to a crawl. As always iTunes is a big offender. Somehow playing MP3s has to be a chore. I think Spotlight is another offender and that's something that Windows XP didn't have.
I use Ubuntu for my Linux desktop, and it has similar problems. There's lag everywhere and it uses a ton of RAM. It's not more usable from a UI perspective. I see a lot of "lateral" progress, although to be fair Ubuntu from 10-15 years ago was pretty bad (definitely less usable than Windows), and they have made progress.
I've got a video: https://twitter.com/capotribu/status/1262021068690259968
Pentium III 547 Mhz, 256 MB RAM, 12GB disk
Sadly don't remember if this video is from cold start or from hibernation
Windows XP only came out in 2001, so if the laptop was from 1999 then it must have been upgraded to Windows XP at some point.
That text based progress bar shows exclusively for resume from hibernate, hence rubatuga saying it's "definitely hibernation".
And yes, XP has exactly the same boot screen. It just goes significantly faster than on 2k, which means most of the time you didn't get to see it.
The text mode loader would only ever appear when resuming from hibernate.
Windows 98 used to take 1-3 minutes to boot on that level of hardware. And if you had any type of network, it was probably even more than 3 minutes. There is a reason hibernation (even at the BIOS level) was all the rage these days.
For better or worse, MS did something on Windows during the XP era that finally made it be able to boot fast (on modern hardware at least). Win2K still has insanely long "delays".
Maybe you misconfigured sth? I remember that I once had something bad in my systemd services, which caused a long hang waiting for nothing and then failing after 1 min timeout. https://askubuntu.com/questions/1234713/why-slow-startup-boo...
I just checked, and I'm 17s from cold boot to Windows 10 login prompt on my Ryzen 2600 desktop with NVMe boot drive. I also checked on one of my 5 year-old HP Zbook G3 (Xeon E3, 64GB RAM, NVMe boot drive), and it was 23s - and that's a corporate machine, loaded with crap.
Objection: nobody had a collection of MP3s in 1995. The computers of 1995 didn't have the power to decompress an audio encoding like MP3 in real-time without an accelerator card. People swapped MIDI arrangements of popular songs, and some folks ripped PCM audio directly to WAV files (where it then took up half their 1GB hard disk, so this was mostly a novelty—unless you were the proud owner of a $2000 1x CD burner to burn the PCM audio back onto.)
MP3s became a big thing around 1999, with better computers and the release of Napster.
WinAmp itself was released in 1997, but wasn't immediately popular because people didn't have music on their computers to play using it. It took off around 1999 as well (which is why everyone remembers Winamp 2, rather than Winamp 1.)
Dang, 1GB back in these days was pretty legit. A 128k MP3 was something like 3-4 megs per song - I'd fit a bunch on a 100 meg Zip disk back in the day and feel like a king.
I remember there was one software decoder that was fast enough in 1995, so I plugged in my Pentium/100 laptop to decode music and pass it uncompressed to my desktop Pentium/90 (which was too slow) over 10base-T!
Changing from one window to another allowed me to see widgets been drawn. It wasn't snappy, but it wasn't a crawl. I could still open a dos window and run qbasic or turboC. I had to be careful: drunk pointers risked stability of the system.
At the time I never had the same thrashing problems I have today running linux. That is why I'm so hopeful about systemd-oomd.
I could keep the machine like this for almost a whole day. A windows 9x machine rarely passed 24 hours uptime with heavy usage without a bluescreen.
I later upgraded the machine to 64mb ram, 2 harddisks, 2 floopy drives and both cd drive bay became occupied: one with a cdrom drive and the other with a cd burner. I then replaced the winmodem with a network card. It was a good machine. The only thing that barred me from using linux on it was the soundcard (AZT-R2316) that had no linux driver. It 2mb s3 video card worked better with linux. By the time the sound drivers arrived it was 2007 and I already had a new linux-only 2006 machine.
On "highly compatible", well, you could run mandrake and send drakconf and KDE to the trash.
I think Audiograbber was the first popular program to rip MP3's from your CD collection. Best I can tell, it was first released as v1.41 in Feb 1997. Winamp came out a couple months later.
And then young people were absolutely listening to their CD's as MP3's on their computers by the summer of 1997 -- I know I was, it was just so much more convenient. And you definitely didn't need to wait for Napster if you were on a college campus with dorms -- you'd find tons of "anonymous" network shares filled with thousands and thousands of tracks in the fall of 1997 for sure.
And the Diamond Rio player came out at the end of 1998... I remember that was huge... all of a sudden your MP3's were portable!
Cheap CDRs were the other big breakthrough. Now people had a reason to download music so they could burn it to listen in their car. But that's well after Win95.
Not computer enthusiasts -- just college students.
I remember being pretty surprised, it wasn't the CS students who had all the ripped music. It was the party kids. The "killer app" for MP3's was the ability to put together all-day and all-night music mixes that could be shuffled instantly. For the first time you could DJ effortlessly.
cdda2wav seems to have existed by 1996.
the "quantum bigfoot" 1GB-and-up HDDs also in 1996 could hold several albums, especially if you ripped at lower rates like 64kBit/s.
I also recall exchanging MP3s on IRC with /dcc send, and am reasonably confident this was 1998 or earlier.
On the other hand, "On July 14, 1995, Karlheinz Brandenburg sent an e-mail announcing the .mp3 file extension had won a poll and announced that the old .bit file extension should no longer be used."
So while literally 1995 is early for people to have a "collection" MP3s on their home machines, the time before everyone had moved on _from_ Windows 95 sure overlapped with the mp3 era, especially for early adopters.
I compiled it on an SGI Indigo running at 30Mhz and it would play up to 256kbps files using 90 percent of CPU.
But this program called "amp"... I'm very curious to see if it could be compiled to run on a recent machine.
I agree with you, but 30Mhz! That is a very low clock for 256kbps decode.
MIPS did very significantly outperform Intel at that time.
http://www.mp3-tech.org/programmer/sources/amp-0_7_6.tgz
Found it.
An Indigo sat in the office for a time playing tunes from an NFS share we would populate. I eventually took it home to do the same thing. I forget how I parted ways with that old box.
I am not inclined to build it at the moment, though it may well make for a great micro controller project one day.
Edit: Fixed: add "frequency = frequency/2;" after the comment /* Set the output frequency */ in file audioIO_Linux.c
Does it still hold up?
Maybe I will build it just for grins myself. Thanks for the tips.
But yes, it uses very little cpu. Maybe less than dosamp.
Maybe I will play your track after the family activities. Edit: Except access denied, lol. No worries.
https://en.wikipedia.org/wiki/WinPlay3 was released in 1995 and requires "only" a 66MHz 486, available in 1992.
What's up with the socket? It looks like there aren't enough pins on the CPU, but they're used for the Pentium Overdrive.
The only way to play mp3s was in linux where I shutdown every possible daemon and ran the decoder at the highest priority.
I used Sonique for a while because it looked amazing - slightly poorer performance, but felt worth it. Later on, Sonique "optimized" for later Pentium releases and it became much too slow on the P100, so I switched back to Winamp.
Also, it still boots. :)
Scour was released in November 1997, and it was released to solve the problem of a whole bunch of college students having their MP3 libraries available via Windows shares, but it was hard to find what you were looking for. Scour fixed that.
So maybe not the general public, but in 1997 someone released a search engine specific to finding MP3s, because people already had large collections.
I was the author of MP3 Fiend around this time.
It was a desktop metasearch engine for MP3s that worked in conjunction with multiple download clients.
It was very successful at the time, only for Napster to come along and eat my lunch.
It got ... called out a lot by the press.
https://i.imgur.com/0hnIG5Y.jpg
LOL, I know exactly what song you mean, but it was called I Want You. We did it in my a cappella group in 1997.
I ran my company's (unofficial) MP3 server in '99, and I spent a LOT of time managing files.
I ultimately went back to college in 2002 which was the right time to catch Apple's first attempt at Home Sharing in iTunes (I forget what it was called), where basically every single Mac on the campus network was running an HTTP server advertising all of its music for download.
I remember them being like 1.5-3 mb each. Not that bad of a download with a 56k
Now I don't even blink over hitting a webpage with multiple megs of assets.
The hacker news favicon is bigger than most webpages of the time :D
I also remember waiting a few minutes to download Winamp, ICQ, and the like. (A dial-up modem would get 4 KB/s down on a good day, so 146 KB was not insignificant for an application.)
Ripping CDs and encoding was definitely a different story.
I recall downloading MOD files back in the day from various BBSs. This is probably pre-95 by a few years. MIDI files were good but you needed a good sound card to play them. I recall drooling over ads of the Gravis Ultrasound (I had a SB Pro).
Sure, the quality is shit by modern standards, but it could easily be played by ordinary computers of the time and took up very little space.
https://youtu.be/0mF9kXZAjsI?t=17
It generally isn't doable. My Amiga 600 without the Vampire or a Prisma Megamix cannot play MP3s of decent quality and the Amiga was far more capable than C64.
yup, that was me. I had to be very selective about what songs I wanted off a small number of CDs.
I seem to remember doing that before ripping and faster CD drives too - instead capturing audio output. That was painful.
I do remember being impressed how mod / tracker based music was so much more efficient on our P5 120Mhz. It was something like 10% of cpu vs 50% for mp3.
3 of us had our Apple t-shirts (from the Apple "Demo Days" program) and lined up at a Sacramento, CA location (I forgot what chain it was) that had an Apple "store in a store" - media (radio) was outside and interviewed us on air about it. We fired up all the Macs, and our friend crashed the big screen demo with the c:\con\con bug.
good times.
When the iPhone launched, my reaction to it was essentially the same as the guy in the computer shop. Boy were we wrong about the impact both would have on the industry!
(* Edit: can't remember if he said Mac or OS/2)
Win98 was much more stable then Win95 and as functional.
That's not really how it works - there's no "console mode" in Linux. On Linux the Desktop environment is built on top of underlying OS interface - that's precisely why you can run it completely headless, or why you can completely switch desktop environments in a few seconds.
If you boot into dos mode on win95 its just stopping the boot process at the start windows step, which is not that different from booting linux but not starting x windows.
If you executed a DOS program within Win 95/3.1 it did not run on the underlying DOS “layer” but in some sort of DOS virtual machine.
So what happened when you booted into “dos mode”? You really booted into DOS like you said without the last step of loading Windows. But loading Windows was not just like startx but more a continuation of the boot process into the Windows Operating system.
telinit 1.
What? IRC was well alive back then and you could have multiple chats in a single IRC client. This claim makes no sense. What ICQ brought was not the idea of multiple chats going on, but a single identifier to connect to people you knew (and that NOT a mobile phone number)
ICQ, and later AIM, made most of this stuff "easy" for the average person. Too bad AIM didn't evolve more... It could've been Slack 20 years earlier.
I wonder what the earliest version of Windows would be that can at least establish a connection to a 2020s website out of the box.
The problem is it isn't going to have SSL versions or cipher suites that are compatible with the modern Internet so it's going to be pretty useless.
[1] https://blog.chocolatey.org/2020/01/remove-support-for-old-t...
You can still open http only websites fine, even in the original internet explorer : https://whynohttps.com/
The good old days. Earlier no body would ever need more than 640K main memory....
You can't compare the very beginning of the internet and what will come from where we are now in the next 10-20 years. Most likely a lot of things will still work in 10 years, because the underlying foundations of the protocols are now standardized and fixed for a while. Same thing with cars from 30 years ago: they can run just fine on the road and with regular gas, because its matured technology.
And honestly where protocols have changed its been very recent. There was a big lull between the 90s and last couple years. TLS 1.3, the most obvious example of something new, was very recent (1.0-1.2 were minor but important improvements to SSL 3, which until recently was still widely supported. TLS1.3 was the biggest change in a long time). Http/2 (and 3!) Are super new, and earlier versions of http which date to the 90s are still universally supported.
I think the take away from that blog post is that the user experience from 95 is not that drastically different from the experience from today. The leap from the 80s UI to 90s was drastic, but it has been getting more stable since. To the point that I would call it stagnated for at least the past 10 years. The web UI did change a lot and, as noticed, it is completely incompatible with browsers from 95. But that's largely because it was catching up with the native apps/ui. I think we are there already, though. So I expect the web to be as stable as the native protocols and graphical toolkits have been for the past 25 years. Probably even more so, since keeping backwards compatibility is much more important in the web.
It’s even more possible today, when a small number of large companies make up for a huge amount of functionality, that we’ll suddenly lose it. GameSpy Arcade was one example of how a single market shift can end entire communities. I fear IoT will be even worse.
XP is barely usable online today, but XP itself was released (2001) before we had anything resembling the modern web. It'd be years before the phrase "Web 2.0" even entered popular lexicon.
Also there was an application, I believe called "gooey" that would automatically create a chatroom for any site you visit. So when you visit a site, there would be a chat for that particular url and people would chat in it.
When I see something that should be tiny being about 250MB, it's usually that they bundled Qt for the GUI. I suppose that's a cost of cross-platform tools. The 10kb executable the author made was using system GUI libraries, a dynamically linked executable on a system with Qt etc already installed would be similar.