Everything going to "the Cloud" is just a backslide into the mainframe era, but with shinier graphics. I feel like the age of micros has ended, killed off by clueless users with flashy phones.
It's just one of the usual cycles. Enough time has gone by that most people have forgotten (or weren't alive) during the last server era. They aren't fresh on what the serious downsides of it are, or why the PC revolution was such a step forward.
But people will be reminded (lots are already beginning to see) and the pendulum will swing back away from using someone else's server. Then after a couple decades, the pendulum will swing back again. Rinse and repeat.
The two main ones are that there's more moving parts to go wrong, and that you're at the mercy of the people running the server. When the server malfunctions, there's nothing you can do but wait for them to fix it, the provider can mess up your account in unpredictable ways leading to unexpected interruptions of service and/or loss of data, and so forth. There are numerous smaller things, too, such as reduced flexibility and such.
Also, communications can and do suffer interruptions, and if you're using a remote server then such interruptions can be problematic.
My point isn't that "the cloud" is a bad model, though. It's that it involves a different set of risks and tradeoffs. For a given use case, "cloud" may be the best solution. For others, it won't be.
That's a snappy take, but it definitely doesn't feel that way to me as someone who did use mainframes a bit and big shared unix servers(not mainframes) quite a lot at points in my career.
The defining characteristic of the mainframe experience for me is extreme rigidity. Upgrades are practically impossible, scheduling time on the mainframe requires lots of forward planning, notice or having a close relationship with a sysop that you can buy drinks for to get jobs to jump the queue.
Cloud on the other hand brings scaling on demand scale up/scale down for bursty workloads, very quick to experiment with different configurations etc. The exact opposite - flexibility is very much the name of the game.
Shared unix servers (not mainframes) were also very different from cloud in this way but also had the difference that you were always in very noticeable contention for CPU and io with other users, whereas on cloud you are almost always on dedicated resources for your use case. I never really felt contention on mainframe workloads because they are so goddamn slow all the time it's pretty much impossible to tell if you're contended or not.
Is it not obvious? It's the same purpose as XCode Cloud and Ubuntu Pro and Office 365; force users to pay recurring fees for basic functionality. This is what service-ification of completely normal software looks like, our penance for not caring enough as consumers.
The "basic functionality" that an OS provides requires constant maintenance from the vendor. You're welcome to use Windows XP if you want, and then you can enjoy not having new device drivers for modern hardware or patches to protect against new attack vectors. Or is Microsoft supposed to patch security vulnerabilities forever, after your one time purchase?
I don't like subscriptions for everything but don't act as though an OS, which must receive continuous updates -- continuous labor from the vendor -- in order to remain secure and thus be usable as an Internet appliance, which is its primary purpose today, is not already a subscription service even if it is not priced as one.
Constant maintenance does not entail running an OS on their servers and RDPing into it. I have a capable machine, I can service it myself. If Microsoft wants to EOL OSes after a one-time purchase, that is their prerogative. There is a distinct difference in expectation between subscription OSes and LTS releases, though. That's why Microsoft can value a Windows license at hundreds of dollars.
OS vendors have a continued interest in supporting my hardware so they can exploit me as a customer. If Microsoft makes too hasty of a move, they know they'll just push their users onto competing platforms that don't charge for 5+ years of support. Stuff like MacOS, ChromeOS or even Linux.
I don't know exactly what the OP was trying to say, but in my experience, Windows runs noticeably slowly and is sluggish compared to Linux on the same hardware. I have a Skylake CPU and 16GB of RAM, too; it's not just a crappy netbook.
I dual boot my laptop with Windows 11 and Linux with GNOME. In Windows, just opening the Settings app takes a few seconds: 1 for the window to actually open, and the rest of the time loading the actual UI (since at first it's just a blank window). Opening applications takes longer. I can open fewer tabs in the web browser before the machine starts to lag. Reading and writing files is slower. (I recall reading that NT made different tradeoffs compared to Linux regarding file I/O and starting processes that made them slower, but the benefits to the tradeoffs they made aren't apparent to me in my usage.)
Running Linux with GNOME (typically said to be bloated compared to other DEs or standalone WMs) is much snappier in all these cases.
I was just thinking about this the other day. If I had to build a new OS built on top of a web interface that controlled an operating system, installed programs, opened files with set programs, shared between programs, and made it all cloud based, what would it look like? How could I do that?
The front would have a set number of UI concepts, theme, interface commands and controls. The back would have data that would appear in the form of drop downs, item list, buttons, images, colors, etc (i think that covers everything).
And then a special renderer would be paired with the interface. This renderer would depict (a VNC?) things like the layout of a webpage (... by firefox?), the structure of a cad drawing, the photoshopping of a 1 GB image, the 10,000 messages in your inbox, all based on the data obtained from the web.
I am not too familiar with GPU stuff so I don't know how much the renderer can partially leverage a local GPU versus a cloud GPU.
(Sidebar: I actually signed up for a trial and somehow Microsoft decided to terminate it not even a day into the trial with zero warning, so I immediately gave up and am annoyed I spent half a day loading software and already starting to use the damn thing.)
The idea of a super thin client in the form of a light laptop with amazing battery life connected to something like this seemed appealing a few years ago.
With everything Apple has delivered with their ARM processors though it seems like you can have your cake and eat it too with compute power, battery life, and a very nice laptop all at once.
Would never be appealing to me in the slightest, having my compute taken away from me and put in someone else's control to siphon away even more data about me, where they have the power to shut my compute off if I violate their TOS, accidentally or not, where I'd lose what little remaining privacy I have left. Of course it'll end up that way in the end eventually, as the enshitification continues; no way would companies and governments not want all that extra data and control.
When AI gets involved it suddenly starts making sense because your laptop's GPU a little baby toy compared to an entire data center running the latest & greatest GPU/AI accelerators.
It might be that every day (e.g. regular people) AI use cases don't scale much beyond 64GB of VRAM though in which case it's only a matter of time before laptops have such capabilities.
Exactly. The naivete of people is astounding. We hear every day about people being deplatformed for no reason at all, and the price they pay for using said platforms.
In a recent sale, I saw a 15" macbook air for $1750 Canadian Dollarydoos (a discount of $250 from the original price of $2000). For t hat money it came with an M2 chip, 512gb SSD, and 8gb of RAM.
There is no alternate timeline of this universe that would make a $1750 or $2000 laptop with 8gb of RAM ok with me in 2023. The prices are high and the upgrade costs for RAM are abusive. Of course, it doesn't use SODIMM modules so you can't upgrade the RAM yourself, you have to buy an even more obscenely priced model from apple.
That being said using thin clients with even less control than existing windows or mac computers have sounds terrible too.
I needed to buy an iMac for my wife's new job, and could only afford the base model iMac M1 24" - 8 GB RAM, small storage, limited Thunderbolt ports, but it's still been the best computer we've ever had, bar none. My wife abused it with huge catalogs built in Adobe InDesign, as well as projects in Illustrator and Photoshop, and the only thing that's held us back is the small storage when Photoshop feels like it doesn't have enough scratch disk. We haven't even noticed the 7 core GPU (vs the 8 core the midrange model gets).
Anectdata aside, you know your workflow better than I do so I won't tell you 8 GB should be enough. It definitely makes it less future proofed for sure.
Totally agree that the current offering from Apple isn’t exactly appealing for everyone, I would imagine other manufacturers will develop products that offer similar specs soon though.
Funnily enough. I live inside Remote Desktop services (have done so for years, and Citrix before that). I have optimized RDP on Linux to some pretty extensive lengths, but I’ve always used iPads as thin clients and this year I decided to use a Mac mini (driving two 5K-ish displays), and the experience is just glorious. Simply glorious, including for video calling (with camera and audio pass-through).
What a great idea to start a video fixed in the bottom right corner so you can't scroll it away. While reading the article it didn't distract me at all. /s
I don't count subscriptions connected with my work because those are my employer's subscriptions, not mine. My employer can, and does, have different preferences about these things than I do.
For people who need Windows for some business requirement, this lets them use less Windows, and have a Mac or Chromebook as their hardware. The result: less Windows.
For people who don't currently use Windows, does a slower, internet-only Windows present a compelling product? Windows doesn't really sell itself (other than as a gaming platform, but we have Nvidia's streaming platform for that), it's a default but if you've not got a business need most people either already have Windows (because the PC was cheap or they play games) or they aren't going to go looking for Windows.
As Windows becomes more and more of a platform for ongoing revenue (read: ads, subscriptions), it's likely to become less attractive for that second group, and that first group will spend less time with it (and see fewer ads).
I get that on-demand cloud machines make sense for many businesses, but I'm not sure that aligns well with my (very naive) understanding of Microsoft's priorities here.
Yep, this is Microsoft's answer to the terrible Citrix user experience, which is a non-trivial component of burnout in healthcare.
Every hospital basically tells their staff, several times a day, via an electronic health record system served over Citrix: we don't trust you, and you time is worthless, fuck you. All in the name of security.
What is so bad about Citrix nowadays? I’m not in the loop anymore but I did Citrix about 15-20 years ago (geez…) and it was working very well and our users liked it. So that was Metaframe XP on Windows Server 2000 and 2003 back then.
Actually, what most hospital IT systems are trying to do is provide “follow me” desktops and an easy to replace endpoint. There are two major private clinic/hospital groups here in Portugal, and at least one of them uses Raspberry Pis and smartcards in doctors’ offices (most of which are interchangeable or need to provide for staff rotation).
It is a pretty sweet system (I noticed no lag as they pulled up my exams and clinical data, including radiology and charts).
A) Consumer Windows PCs and laptops have been getting cheaper and cheaper. You can get a Windows laptop for $200 at any Walmart. That's $200 in 2023 dollars versus for example $1000 for desktops in 1997 dollars. Microsoft's money off of Windows consumer licenses has to be declining.
B) Windows may be dying but Active Directory isn't, plus everything moving to cloud.
For the business side, it started with Microsoft's hosted email service - a subscription service many, many businesses saw value in using over hosting their own email. What's happening is the amount of things hostable is growing until Microsoft basically has Active Directory completely in the cloud. Then businesses can run with even less IT staff than they do now, but still have all the access control that Active Directory offers, even among non-Microsoft products that support SSO.
>Windows doesn't really sell itself (other than as a gaming platform, but we have Nvidia's streaming platform for that)
And the Steam Deck shows that you can run nearly any Windows-based game locally on a machine not running Windows at all (Wine/Proton on Linux, but a Steam Deck user doesn't have to worry about the details; it just works).
This article is nonsense. Microsoft intentionally made it hard and expensive for innovative companies to try and deliver Windows as a service through a browser for years. They didn’t want to hurt their lucrative OEM licensing business. This made it a thing for enterprise customers.
What do you mean? OEM licensing isn’t the same as enterprise volume licensing. And the latter is per seat/user anyway. Plenty of projects deliver remote rdp session over a browser fine. It even works better than VNC in the browser.
What if I just run a bunch of cryptominers on my Windows 365 instance? Either way this is gross and I guarantee people will just accept that. To unintentionally quote Deus Ex, Some people don't see the danger of indiscriminate surveillance.
55 comments
[ 3.1 ms ] story [ 106 ms ] threadBut people will be reminded (lots are already beginning to see) and the pendulum will swing back away from using someone else's server. Then after a couple decades, the pendulum will swing back again. Rinse and repeat.
Also, communications can and do suffer interruptions, and if you're using a remote server then such interruptions can be problematic.
My point isn't that "the cloud" is a bad model, though. It's that it involves a different set of risks and tradeoffs. For a given use case, "cloud" may be the best solution. For others, it won't be.
Perhaps slightly dated, but still relevant and hilarious: The Bastard Operator from Hell
http://bofh.bjash.com/
The defining characteristic of the mainframe experience for me is extreme rigidity. Upgrades are practically impossible, scheduling time on the mainframe requires lots of forward planning, notice or having a close relationship with a sysop that you can buy drinks for to get jobs to jump the queue.
Cloud on the other hand brings scaling on demand scale up/scale down for bursty workloads, very quick to experiment with different configurations etc. The exact opposite - flexibility is very much the name of the game.
Shared unix servers (not mainframes) were also very different from cloud in this way but also had the difference that you were always in very noticeable contention for CPU and io with other users, whereas on cloud you are almost always on dedicated resources for your use case. I never really felt contention on mainframe workloads because they are so goddamn slow all the time it's pretty much impossible to tell if you're contended or not.
To do what ? Running on bare hardware is a challenge for Windows.
Sounds simple to me /s
Think of all of the telemetry data possible this way!
Is it not obvious? It's the same purpose as XCode Cloud and Ubuntu Pro and Office 365; force users to pay recurring fees for basic functionality. This is what service-ification of completely normal software looks like, our penance for not caring enough as consumers.
I don't like subscriptions for everything but don't act as though an OS, which must receive continuous updates -- continuous labor from the vendor -- in order to remain secure and thus be usable as an Internet appliance, which is its primary purpose today, is not already a subscription service even if it is not priced as one.
OS vendors have a continued interest in supporting my hardware so they can exploit me as a customer. If Microsoft makes too hasty of a move, they know they'll just push their users onto competing platforms that don't charge for 5+ years of support. Stuff like MacOS, ChromeOS or even Linux.
What do you mean?
I dual boot my laptop with Windows 11 and Linux with GNOME. In Windows, just opening the Settings app takes a few seconds: 1 for the window to actually open, and the rest of the time loading the actual UI (since at first it's just a blank window). Opening applications takes longer. I can open fewer tabs in the web browser before the machine starts to lag. Reading and writing files is slower. (I recall reading that NT made different tradeoffs compared to Linux regarding file I/O and starting processes that made them slower, but the benefits to the tradeoffs they made aren't apparent to me in my usage.)
Running Linux with GNOME (typically said to be bloated compared to other DEs or standalone WMs) is much snappier in all these cases.
The front would have a set number of UI concepts, theme, interface commands and controls. The back would have data that would appear in the form of drop downs, item list, buttons, images, colors, etc (i think that covers everything).
And then a special renderer would be paired with the interface. This renderer would depict (a VNC?) things like the layout of a webpage (... by firefox?), the structure of a cad drawing, the photoshopping of a 1 GB image, the 10,000 messages in your inbox, all based on the data obtained from the web.
I am not too familiar with GPU stuff so I don't know how much the renderer can partially leverage a local GPU versus a cloud GPU.
https://www.microsoft.com/en-us/windows-365/all-pricing
(Sidebar: I actually signed up for a trial and somehow Microsoft decided to terminate it not even a day into the trial with zero warning, so I immediately gave up and am annoyed I spent half a day loading software and already starting to use the damn thing.)
With everything Apple has delivered with their ARM processors though it seems like you can have your cake and eat it too with compute power, battery life, and a very nice laptop all at once.
It might be that every day (e.g. regular people) AI use cases don't scale much beyond 64GB of VRAM though in which case it's only a matter of time before laptops have such capabilities.
Still a lot of unknowns though.
Email, photos, notes, relationships, documents, professional work...every part of their life is already in the cloud.
There is no alternate timeline of this universe that would make a $1750 or $2000 laptop with 8gb of RAM ok with me in 2023. The prices are high and the upgrade costs for RAM are abusive. Of course, it doesn't use SODIMM modules so you can't upgrade the RAM yourself, you have to buy an even more obscenely priced model from apple.
That being said using thin clients with even less control than existing windows or mac computers have sounds terrible too.
Anectdata aside, you know your workflow better than I do so I won't tell you 8 GB should be enough. It definitely makes it less future proofed for sure.
For people who need Windows for some business requirement, this lets them use less Windows, and have a Mac or Chromebook as their hardware. The result: less Windows.
For people who don't currently use Windows, does a slower, internet-only Windows present a compelling product? Windows doesn't really sell itself (other than as a gaming platform, but we have Nvidia's streaming platform for that), it's a default but if you've not got a business need most people either already have Windows (because the PC was cheap or they play games) or they aren't going to go looking for Windows.
As Windows becomes more and more of a platform for ongoing revenue (read: ads, subscriptions), it's likely to become less attractive for that second group, and that first group will spend less time with it (and see fewer ads).
I get that on-demand cloud machines make sense for many businesses, but I'm not sure that aligns well with my (very naive) understanding of Microsoft's priorities here.
Every hospital basically tells their staff, several times a day, via an electronic health record system served over Citrix: we don't trust you, and you time is worthless, fuck you. All in the name of security.
It is a pretty sweet system (I noticed no lag as they pulled up my exams and clinical data, including radiology and charts).
It's the fucking EHRs
The people who write the specs are so far from the end-users the result is mind-bogglingly terrible for said end-users
That the same people who obsess about their IDEs subject their ultimate end users to these monstrosities of inefficiency is some kind of sick joke
A) Consumer Windows PCs and laptops have been getting cheaper and cheaper. You can get a Windows laptop for $200 at any Walmart. That's $200 in 2023 dollars versus for example $1000 for desktops in 1997 dollars. Microsoft's money off of Windows consumer licenses has to be declining.
B) Windows may be dying but Active Directory isn't, plus everything moving to cloud.
For the business side, it started with Microsoft's hosted email service - a subscription service many, many businesses saw value in using over hosting their own email. What's happening is the amount of things hostable is growing until Microsoft basically has Active Directory completely in the cloud. Then businesses can run with even less IT staff than they do now, but still have all the access control that Active Directory offers, even among non-Microsoft products that support SSO.
And the Steam Deck shows that you can run nearly any Windows-based game locally on a machine not running Windows at all (Wine/Proton on Linux, but a Steam Deck user doesn't have to worry about the details; it just works).
It's to get enterprises and consumers to pay for Windows twice:
1) once for the Windows that runs on hardware and controls devices (traditional Windows)
2) continually for Windows and related services that run in the cloud (Windows 365, CoPilot, etc.) on a subscription basis
What was Microsoft doing to make it hard exactly?