41 comments

[ 2.8 ms ] story [ 84.6 ms ] thread
To clarify, the (1993) does indeed signify the release of this in 1993. This isn't a historical software release (this folded in many improvements from external contributors because it was open source).

Maybe I'm dumb but I did need to do a double take.

it's a pretty fun exercise to download the earliest mosaic browser you can find and get it to build on your current platform. Then to see what sites actually still work. The last time I did this, Dennis Ritchie's site loaded the best.
Made this experiment a while ago too. Many pages, even plain simple ones, fail because it doesn‘t support the HTTP host header and therefore name based virtual hosts.

It can only display one site per IP and port.

jwz has a proxy you can run to inject the Host header for old browsers https://www.jwz.org/hacks/http10proxy.pl
Is this the HN equivalent of the old "goatse" links? Or do some people on HN not realizes that jwz substitutes a page with an explicit graphic image when linked to on HN?

Would be nice to see an HN update that converts jwz links to non-clickable.

Ugh, sorry my bad. I always open links from comments in new tabs and Safari doesn't send a Referer in that case so I had completely forgotten that he does that
Yes, @dang, is it possible to ban jwz links?
Better yet, special-case them to add rel="noreferrer". If jwz wants to break a fundamental part of how the web works, HN can break it right back.
Last time I tried this out on my NeXT I found that pervasive use of HTTPS makes it pretty hard to find things to try out.
I like how WWW is one protocol amongst many:

> a wide variety of networked information sources, including Gopher, WAIS, World Wide Web, NNTP/Usenet news, Techinfo, FTP, local filesystems, Archie, finger, Hyper-G, HyTelnet, TeXinfo, telnet, tn3270, and more.

So Mosaic was also a local file browser? I always thought this stupid idea to combine web and local file browsing into one app was invented by Microsoft and then later adopted by KDE (Konqueror)...
It is a nominal feature that exists in part to supplement the file: protocol for viewing local web pages. The UI is invariably basic, typically identical to the browser's FTP interface.

Many (most?) web browsers support this, even today.

In the early days web sites would often be down, or you might need to copy pages or a whole site to disk for access offline, especially on laptops. Even in the early 2000s I used to take dumps of HTML documentation with me on disk or thumb drive, because even if I had internet access at a hotel or customer site I wouldn't have access back to the company network. It was an essential feature.
I constantly use this feature, even today. I regularly use Dash, and when I’m working on a personal project I like to use SiteSucker or safaris WebArchive save format to collect the docs, information and blog posts I need.

With this, I can then turn off my internet and dive deep into the task. It’s how I learn!

About the only missing thing I have is an offline cache for language package managers — I use so many that it’s a bit difficult. I should really go look for some Docker images that have already tackled it...

At the time it was just one among many. I first encountered it using telnet info.cern.ch, which was a line mode browser. There wasn't much to see other than links to other web servers. Usenet was much more important and gopher had more content. But when I first saw a web page on Mosaic, with pictures and rendered text, it was mind blowing.
Love the first reply to the release announcement:

Hi Marc,

thanks for releasing this. I found it to be a useful little program, however the Look&Feel seems a bit dated.

Best regards, Manuel

Note the time stamp. That was posted today, not in 1993.
—Manuel, the Troll
I wonder how did they take Mosaic from NCSA to found the business (later to be renamed to Netscape)? Did they really reimplement the browser, or maybe they had rights?
If you want to know all the intrinsic little details on Mosaic, NCSA and Netscape I would recommend the book Weaving the Web. It is written by Sir Tim Berners-Lee back in 1999 and is very close on the action.
A number of companies licensed the rights to that code (it's in their credits as "Spyglass, Inc."). I just assumed that Netscape had a relationship with NCSA, but the browser that kept the "Spyglass, Inc." credit longest was actually IE, as far as I know.
They reimplemented it. Jim Clark mandated it IIRC
MCOM (later Netscape) and NCSA were definitely not on friendly terms. There was a lawsuit against MCOM at which point they changed their name.
> Added new resource, gethostbynameIsEvil, for Sun's that coredump when gethostbyname() is called to try to find out what their own names are

Solid OS engineering right there

My recollection is that the DECs were generally more stable, but my recollection also has 25 years or more of cruft.
Further down the Changelog:

> Fixed mysterious stupid coredump that only hits Suns.

Man, I miss my Sun admin days. Always exciting.

When I went to university in 2003 they had nothing but Suns. Had a lot of "fun" trying to compile modern software like Gaim to run on that, while keeping within my measly quota. There was a repository where everyone would share the paths to the dependencies they had built so you wouldn't have to waste your own quota.

Was kinda fun running Microsoft Internet Explorer for UNIX though.

Compiled this on Linux with very minor changes (which were limited to Makefile edits), and still works - even on Wayland, via Xwayland!

Unfortunately it doesn't support HTTP 1.0 so no one is willing to talk to me. I was able to get it to render local HTML files, though:

https://sr.ht/ULoK.png

Update: gopher and FTP work:

https://sr.ht/MDUE.png

https://sr.ht/0E09.png

Telnet SIGABRTS and gives me an email address to send the core dump to (highly tempting). If anyone knows of a working finger server, I'll give it a shot.

Update 2: mailed a core dump to that address, it bounced :( reached out to postmaster@ to see where they went.

It looks like the gopher support is old enough that it doesn't know about the (non-standard, but everyone uses it) 'i' item type for informational menu items, which is why that entire page, including the ASCII art and intro text, is one big pile of links.