7 ms·

Ask HN: How did open source work before GitHub?

43 points by pvsukale3 9y ago

How would you publish maintain projects? How would issues be resolved? It will be nice if mention the era in the answer . like 80's 90's 2000's

whatnotests 9y ago
Before GitHub you had Google Code. Before that was SourceForge.
- srinathrajaram 9y ago
  Can confirm sourceforge was huge
  - zhte415 9y ago
    Another vote for Sourceforge; even today it contains a lot of stuff.
    - cocoa19 9y ago
      Sourceforge got desperate for money and they started modifying installation packages to contain crapware, an example is FileZilla. I try to stay away from sourceforge since then.
      - mdekkers 9y ago
        Sourceforge got desperate for money and they started modifying installation packages to contain crapware That is only a relatively recent development.
        protomyth 9y ago
        And one they have stated here on HN that the new owners are ending. Judging by the comments everytime they are mentioned, they need to do a lot more aggressive PR.
        loganabbott 9y ago
        Trust me, we are trying.
        dragandj 9y ago
        They can do whatever they want, many people won't touch them for the rest of their lives, I'm among them.
        protomyth 9y ago
        How is that a productive attitude in life towards anything? The people who did the deed are long gone, but the people who want to restore and advance SourceForge are SOL?
        dragandj 9y ago
        Because there is competition that have never done anything bad like SF, that offers better products. Would you buy your groceries at Al Capone's, while there's a grandpa's groceries just right next door that offers better, cheaper, and more versatile stuff?
        protomyth 9y ago
        SourceForge isn't owned by the same people anymore, so your example is rather bogus.
        quadrangle 9y ago
        It's not recent at all, it's just something that happened since GitHub first launched, but they had that issue for years. And the recent development is having new owners who stopped that bullshit.
        mdekkers 9y ago
        relatively recent - they have been around since 1999, and the scamware stuff didn't start until 2013.
        MilnerRoute 9y ago
        SourceForge now scans all projects for malware and displays warnings on downloads (sourceforge.net) 3 points by loganabbott 367 days ago https://news.ycombinator.com/submitted?id=loganabbott https://news.ycombinator.com/submitted?id=loganabbott
        protomyth 9y ago
        I think mdekkers meant it was a recent development in the history of SourceForge and was not true for most of its existence. It has been rectified.
      - quadrangle 9y ago
        SF has new owners, all the crapware and other worst practices have been gone for months.
- stevekemp 9y ago
  And to help you find the projects, and their new releases, back in the day, there were directories such as http://freshmeat.net/ http://freshmeat.net/ (Something I think of re-creating now and again, but maybe with distributions containing packages people don't need such a thing any more?)
  - rkeene2 9y ago
    Freshcode still exists as a reincarnation at http://freshcode.club/ http://freshcode.club/
coreyp_1 9y ago
Just sort by year: https://en.wikipedia.org/wiki/Comparison_of_source_code_hosting_facilities https://en.wikipedia.org/wiki/Comparison_of_source_code_host... Before then, code was often on people's private websites, or FTP servers, with communication over BBS, email, and IRC.
blacksqr 9y ago
Before the web there was usenet. People would assemble and curate lists of links and post them to newsgroups dedicated to particular languages or fields of interest. People would assemble CDs stuffed with code, you'd send them a few bucks and they'd send you a copy by mail. In the 1980s you'd buy computer magazines and check the classifieds. For a few bucks you'd get a floppy disk mailed to you. Netscape did a lot to move project management to the web. Bugzilla was written in the mid-1990s. Before that all bug reporting and issue tracking was done via email. If you were good you'd email a patch to the author to fix a bug or add a feature, just as people still do with Linus. He wrote git initially as an improved tool to merge patches received by email.
- Isamu 9y ago
  Walnut Creek sold a CD-ROM that gathered up a lot of usenet source lists. To those of us who had limited access to the internet this got us pretty far. I remember writing some extensions to usenet software and emailing the author. That opened up a whole world I never knew existed.
brianmurphy 9y ago
Some projects would use older version control systems like cvs. Others would manage development through email lists. Everyone was good at using the diff & patch commands. Mail clients like mutt were great at searching and sorting email. Releases would go up to FTP servers. In larger projects, there might have been multiple trees maintained. For example, get your patch into Andrew or Alan's tree before Linus would take it.
db48x 9y ago
The only thing that Github's pull request system adds that wasn't being done before is single-click merging. Everything else was there. You would email a patch to the author, or to the mailing list, and discuss the merits of the patch there. Then someone with access to the repository would commit it (I use the term generically; version control was less common back then). The diff program we all use today has it's origins in the 70s, although the unified diff format that we almost always use was created in the 80s. Email was invented in the 60s, UUCP in the 70s, and Usenet in 1980. Version control was invented in the 70s with SCCS. Web-based bug trackers such as Bugzilla originated in the 90s, as did continuous integration software like Tinderbox. Of all of these, Git represents the biggest improvement over it's predecessors, while email is pretty much still just email. The diff program hasn't changed since the 80s. Bug trackers and other such tools haven't changed dramatically either, although a lot more attention is paid to UX and polish these days. I think the largest change is the wide availability of these tools today. Everyone knows about them, and it's seen as aberrant to not use version control or bug tracking.
yladiz 9y ago
As other commenters mentioned, you used Google Code, Sourceforge, and before that you'd email a patch to an email list, specific bug tracker for the project, or to the maintainer. Probably the biggest changes that Github added was the social aspect, and much improved UI compared to most previous tools. It made it much easier to search, browse code in the browser, and add requests to merge features or bug fixes.
falcolas 9y ago
Mailing lists and Newsgroups (usenet) were both quite popular. Before the internet, it was over dialup BBS systems and CDs/floppy disks. Just remember that Git was designed around the workflow of the Linux kernel originally. That's why there are commands like "git format-patch" and "git am" which have help messages like "Prepare patches for e-mail submission" and "Apply a series of patches from a mailbox". It was also not that uncommon to see code patches printed in magazines or snail-mail letters.
- tyingq 9y ago
  This was my experience as well. In fact, source code control tools weren't that high on the list of success factors. Communication and distribution were more important. "Pull requests" were often just a tarball of diffs against a specific version, sent via email, and discussed on usenet. The code owner had to deal with the merge issues if the tarball wasn't built against "trunk".
mdekkers 9y ago
sourceforge
protomyth 9y ago
OpenBSD with anoncvs started in the 90's with showing a live cvs to the world. That was a nice change from release dumps.
user5994461 9y ago
It's very simple: "Just use sourceforge". https://sourceforge.net/ https://sourceforge.net/ GitHub didn't invent anything. They only made a better UI. SourceForge was more oriented toward software users. It handles the distribution of binaries. Imagine a time when 90% of software sites would redirect you to sourceforge to download them. Distribution was first. The repo and the issue tracker were in a secondary tab. GitHub is oriented toward developers. It shows you the repo as the main page. You can't distribute binary releases (well, it was added recently). GitHub took over by attracting developers right at the start of the project. They need a repo, here it is. UI is nice, site works. The new projects eventually become well known and carry the reputation. SourceForge went downhill slowly with their shady practices. Ads, confusing pages and software repackaged with adware.
- brians 9y ago
  I remember being so confused by SF in its first years. Why would I want this, instead of CVS in AFS or over SSH? The idea that most software developers wouldn't have routable workstations, wouldn't have servers they controlled---that we wouldn't keep running networked file systems over the Internet---totally not obvious. The SF people foresaw a need in a changing world and got in front of it.
  - pjc50 9y ago
    When and where were you that use of AFS was normal?
    - scolson 9y ago
      Not the commenter, but guessing AFS = andrew file system. This has been used for decades by a lot of the universities who were original internet consumers (MIT, Univ of Michigan, Carnegie Mellon, etc). In fact, it is still used by a lot of these universities and research institutions for data sharing and distributed computing, afaik.
      - danieldk 9y ago
hprotagonist 9y ago
It's worth noting that kernel development is still git + emailing patches around. It's too big for github/gitlab.
- algesten 9y ago
  What are you talking about?! https://github.com/torvalds/linux https://github.com/torvalds/linux Linus just don't agree with GitHub's pull-requests, so he doesn't accept such that way.
  - quadrangle 9y ago
    You know about the kernel but not how to use your compose key‽ The fact that the kernel is mirrored on GitHub doesn't mean they use it in any sense like the way most projects use it. The real point is that the way Linux is developed has no relation to that GitHub mirror, so it's an example of open source development outside of GitHub.
  - cyphar 9y ago
    I don't think anyone is claiming that GitHub can't host a mirror of a big git repo. That's not what parent is claiming is "too big". The problem is that the tooling for reviewing changes on GitHub doesn't scale and also doesn't match the model that Linux uses (deputies). Hell, I would argue large projects like Kubernetes and Rust are hitting these same issues. Also, that is just a mirror. The real hosting is on https://git.kernel.org/ https://git.kernel.org/.
    - algesten 9y ago
      Ok. I read it as "too big" size wise. Of course I know it's a mirror.
holydude 9y ago
Many people commenting here say that github did not bring anything new to the game. While that is partially correct we know that bringing only new things does not matter. What matters is slick,seamless polishing of existing ideas and execution. There are many examples of just reinventing the wheel and making it more beatiful and usable. To your original question: sourceforge was huge, google code, cvs, bazillion of local "downloads.com" etc
- pluma 9y ago
  I think it's incredibly naive to underestimate how much value GitHub provided by shifting the focus from users to developers and putting contribution front and center. SourceForge was primarily a way to distribute software. While most software would be under some form of open source license you'd often also find plenty of projects that either weren't really open source or hosted their source code elsewhere. And when the source was on SF, the web interface for browsing it was unintuitive and unhelpful. The most intuitive use of SF for non-maintainers was downloading releases and leaving comments. There was a separate space for filing issues but it was pretty much just another forum. And let's not forget that contributing code also meant you had to have experience with using SVN (or CVS, I think?) because learning that at the time felt even more daunting than git does now. Before THAT you were mostly confined to sending patches to newsgroups or mailing lists as far as I recall. I open sourced a single project on SF and beyond that never really got into open source until GitHub made it painless and straightforward -- I doubt I'm the only one.
  - k__ 9y ago
    So true. Coming from times when I had to download my PHP libs from SF and trying to get some fixes landed, everything's much different and better now. I get everything via a simple npm install and landing PRs has never been easier.
ing33k 9y ago
Others have pointed at sourceforge , google code . There was also Freshmeat ( renamed to Freecode ) https://en.wikipedia.org/wiki/Freecode https://en.wikipedia.org/wiki/Freecode
roadbeats 9y ago
See suckless.org for a good example. Personally, I used to put them under my personal website, seperating each to a subdomain. If you'd like to see source code, you'd download it.
pjc50 9y ago
Kids these days :) Richard M Stallman will tell you of the days when people mailed tapes back and forth, but that was in the 70s. A key development was Larry Wall writing "patch" in the early 80s, which he posted to USENET. That made it much easier to start mailing changesets to the maintainers of programs. This workflow continued basically until Linux wrote "git" in the late 2000s. The idea of the pull request formalises the process of emailing Linus. In the late 90s and early 2000s, most projects had their own CVS server and mailing list. Mozilla decided they would write Bugzilla to keep track of bugs. Sourceforge grew up as an ad-supported hosting and distribution server for CVS (and later SVN) repositories.
- hk__2 9y ago
  > until Linux wrote "git" in the late 2000s typo: it’s "Linus".
- anjbe 9y ago
  Worth mentioning also that in the early through mid 90s, projects hosted their own version control servers, but you needed an account on the development machine to actually have access to the repo. Most people only had access to release tarballs, not the full history, and couldn’t (e.g.) make diffs against CVS head. OpenBSD developed anonymous CVS, which gave anybody the ability to get the full project history even if they weren’t a developer. (That’s actually where the “Open” in OpenBSD comes from—it refers not to the license, but the development process.) Here’s the anoncvs paper: https://www.openbsd.org/papers/anoncvs-paper.pdf https://www.openbsd.org/papers/anoncvs-paper.pdf OpenBSD’s latest release song was actually about this topic: https://www.openbsd.org/lyrics.html#61 https://www.openbsd.org/lyrics.html#61
- Turing_Machine 9y ago
  "Richard M Stallman will tell you of the days when people mailed tapes back and forth" There's a story of an AT&T source tape being left on a street corner by an anonymous donor in the early days of (I think) BSD.
- luckydude 9y ago
  Larry Wall's patch was a huge step forward. In case you don't know what this is, it takes the output of diff and tries to apply those diffs to your source files. If diffed baseline and your baseline are the same then the patch will apply cleanly. If they are close the patch may apply cleanly (though it may have some messages like "offset 3 lines" which tells you that it found the place to patch 3 lines away from where it thought it was). There was a big period of time where the Linux kernel work was all diff/patch based. So the answer was (a) a mailing list that everyone was one, (b) some way of knowing a baseline (you could just have a baseline tree and a copy of that and you diff -Nur the two trees), (c) email the patch. All that worked but it didn't scale which is why Linus switched to BitKeeper and then later Git.
catwell 9y ago
I started being involved with Open Source in the early 00s. For source code, people typically used CVS or (later) SVN. The repos were hosted by platforms like SourceForge, berliOS or Savannah, or on the projects' own servers. Releases were tarballs hosted on servers. End users sometimes got software by downloading and compiling those, but there were already package managers and repositories. Before most people had Internet at home, at least in France, you could buy CDs with Linux packages at the press store. Communication between developers typically happened on IRC for casual chat and through mailing lists (hosted with mailman) for serious stuff. Sometimes there were web forums to communicate with users. And conferences, of course. A lot of projects were developed by people who were physically close, for instance people who met at the same Linux Users Group. Some projects used bugtrackers actively, some did not. It was mostly Bugzilla or Trac. That did not really change too much with GitHub. The main difference in my mind is that forking was usually regarded as something bad. You would not fork a project and then send pull requests to fix issues. For a simple fix, you would generate a patch and send it to the project's mailing list, where it would be discussed and eventually checked in by one of the owners. If you wanted to really engage with the project, you would become active in the community and eventually get commit rights. I think what led to that change of model is: cheap branching in Git, GitHub as a platform with users account, and maybe Continuous Integration making it easier to validate patches.
jedisct1 9y ago
Sunsite. And new releases were announced on Freshmeat.
kem 9y ago
God this more than anything before has made me feel old. Is GitHub that strongly associated with open source that alternate models for open source development are a mystery? Somehow that's really troubling to me in terms of the open source paradigm, even though I generally see GitHub as a very positive thing. I don't mean that as a criticism of the poster, just saying that it feels like maybe there's some room for more diversity and competition. I think others have given pretty good descriptions of pre-GitHub models, so I'll leave that to them.
geraldcombs 9y ago
Self-hosting. SourceForge was certainly popular at one time, but it didn't own the market, certainly not to the extent that GitHub currently does. Many projects were simply run on whatever hardware and connectivity the project could scrounge up. For a project I founded I initially bought a used SPARCstation on EBay and installed CVS, Apache, Mailman, Bugzilla, etc. I traded consulting time with various local ISPs in exchange for bandwidth.
herval 9y ago
SourceForge. Before that, webpages in a "ring". Before that, Usenet/BBS forums. SourceForge, specifically, is basically the same as GitHub
tzs 9y ago
In addition to the repository hosts already mentioned, there were also some specialized services for open source for a particular type or class of use. A good example was CurseForge.com. They provided free SVN, Git, and Mercurial hosting for add-ons for several popular games, along with per add-on issue tracking and discussion. Add-on authors could host at other, general purpose, hosting sites, but there was an advantage to using CurseForge. Curse.com had an application called the Curse Client for end users. The Curse Client was an add-on manager. It integrated with CurseForge, allowing you to handle searching for, downloading, installing, and updating add-ons from CurseForge right in the Curse Client GUI. Curse and CurseForge are still around, and support a variety of games. I'm not sure what the deal is with repository hosting there now. I saw some documentation that suggested you are supposed to use an external repository and provide a link to it as part of your CurseForge project settings, but I also see a lot of projects whose repository is on curseforge.com.
chrisgoman 9y ago
Self-hosted Trac (without the "k"). It was usually trac.domain.com
payne92 9y ago
zipped source tarballs w/ anonymous FTP downloads. Then you emailed diff patches to maintainers and prayed.
NoCanDo 9y ago
Sourceforge or SVNs. Github didn't revolutionize the open source landscape. They just gave everybody an easier-to-use GUI.
skyzyx 9y ago
I managed SimplePie using a self-hosted SVN repository for the first 5 years of its life. Contributions exploded AFTER moving to GitHub. The AWS SDK for PHP started life as a project called Tarzan on Google Code. GitHub made code feel more accessible. It made contributions easier. It improved the visibility inside the project.
oceanghost 9y ago
Hi, I was one of the original contributors to Gentoo in the late 90s (back then it was called Enoch). We had a private FTP site we stored source code and packages at, and we all downloaded regular snapshots of the system. I would archive milestones to CD-R out of a sense of duty+fear. It sounds primitive, but at the time source control was something only wealthy corporations had.

43 points by pvsukale3 9y ago

How would you publish maintain projects? How would issues be resolved? It will be nice if mention the era in the answer . like 80's 90's 2000's

whatnotests 9y ago

Before GitHub you had Google Code. Before that was SourceForge.

srinathrajaram 9y ago
Can confirm sourceforge was huge
- zhte415 9y ago
  Another vote for Sourceforge; even today it contains a lot of stuff.
  - cocoa19 9y ago
    Sourceforge got desperate for money and they started modifying installation packages to contain crapware, an example is FileZilla. I try to stay away from sourceforge since then.
    - mdekkers 9y ago
      Sourceforge got desperate for money and they started modifying installation packages to contain crapware That is only a relatively recent development.
      - protomyth 9y ago
        And one they have stated here on HN that the new owners are ending. Judging by the comments everytime they are mentioned, they need to do a lot more aggressive PR.
        loganabbott 9y ago
        Trust me, we are trying.
        dragandj 9y ago
        They can do whatever they want, many people won't touch them for the rest of their lives, I'm among them.
        protomyth 9y ago
        How is that a productive attitude in life towards anything? The people who did the deed are long gone, but the people who want to restore and advance SourceForge are SOL?
        dragandj 9y ago
        Because there is competition that have never done anything bad like SF, that offers better products. Would you buy your groceries at Al Capone's, while there's a grandpa's groceries just right next door that offers better, cheaper, and more versatile stuff?
        protomyth 9y ago
        SourceForge isn't owned by the same people anymore, so your example is rather bogus.
      - quadrangle 9y ago
        It's not recent at all, it's just something that happened since GitHub first launched, but they had that issue for years. And the recent development is having new owners who stopped that bullshit.
        mdekkers 9y ago
        relatively recent - they have been around since 1999, and the scamware stuff didn't start until 2013.
      - MilnerRoute 9y ago
        SourceForge now scans all projects for malware and displays warnings on downloads (sourceforge.net) 3 points by loganabbott 367 days ago https://news.ycombinator.com/submitted?id=loganabbott https://news.ycombinator.com/submitted?id=loganabbott
        protomyth 9y ago
        I think mdekkers meant it was a recent development in the history of SourceForge and was not true for most of its existence. It has been rectified.
    - quadrangle 9y ago
      SF has new owners, all the crapware and other worst practices have been gone for months.
stevekemp 9y ago
And to help you find the projects, and their new releases, back in the day, there were directories such as http://freshmeat.net/ http://freshmeat.net/ (Something I think of re-creating now and again, but maybe with distributions containing packages people don't need such a thing any more?)
- rkeene2 9y ago
  Freshcode still exists as a reincarnation at http://freshcode.club/ http://freshcode.club/

coreyp_1 9y ago

Just sort by year: https://en.wikipedia.org/wiki/Comparison_of_source_code_hosting_facilities https://en.wikipedia.org/wiki/Comparison_of_source_code_host... Before then, code was often on people's private websites, or FTP servers, with communication over BBS, email, and IRC.

blacksqr 9y ago

Before the web there was usenet. People would assemble and curate lists of links and post them to newsgroups dedicated to particular languages or fields of interest. People would assemble CDs stuffed with code, you'd send them a few bucks and they'd send you a copy by mail. In the 1980s you'd buy computer magazines and check the classifieds. For a few bucks you'd get a floppy disk mailed to you. Netscape did a lot to move project management to the web. Bugzilla was written in the mid-1990s. Before that all bug reporting and issue tracking was done via email. If you were good you'd email a patch to the author to fix a bug or add a feature, just as people still do with Linus. He wrote git initially as an improved tool to merge patches received by email.

Isamu 9y ago
Walnut Creek sold a CD-ROM that gathered up a lot of usenet source lists. To those of us who had limited access to the internet this got us pretty far. I remember writing some extensions to usenet software and emailing the author. That opened up a whole world I never knew existed.

brianmurphy 9y ago

Some projects would use older version control systems like cvs. Others would manage development through email lists. Everyone was good at using the diff & patch commands. Mail clients like mutt were great at searching and sorting email. Releases would go up to FTP servers. In larger projects, there might have been multiple trees maintained. For example, get your patch into Andrew or Alan's tree before Linus would take it.

db48x 9y ago

The only thing that Github's pull request system adds that wasn't being done before is single-click merging. Everything else was there. You would email a patch to the author, or to the mailing list, and discuss the merits of the patch there. Then someone with access to the repository would commit it (I use the term generically; version control was less common back then). The diff program we all use today has it's origins in the 70s, although the unified diff format that we almost always use was created in the 80s. Email was invented in the 60s, UUCP in the 70s, and Usenet in 1980. Version control was invented in the 70s with SCCS. Web-based bug trackers such as Bugzilla originated in the 90s, as did continuous integration software like Tinderbox. Of all of these, Git represents the biggest improvement over it's predecessors, while email is pretty much still just email. The diff program hasn't changed since the 80s. Bug trackers and other such tools haven't changed dramatically either, although a lot more attention is paid to UX and polish these days. I think the largest change is the wide availability of these tools today. Everyone knows about them, and it's seen as aberrant to not use version control or bug tracking.

yladiz 9y ago

As other commenters mentioned, you used Google Code, Sourceforge, and before that you'd email a patch to an email list, specific bug tracker for the project, or to the maintainer. Probably the biggest changes that Github added was the social aspect, and much improved UI compared to most previous tools. It made it much easier to search, browse code in the browser, and add requests to merge features or bug fixes.

falcolas 9y ago

Mailing lists and Newsgroups (usenet) were both quite popular. Before the internet, it was over dialup BBS systems and CDs/floppy disks. Just remember that Git was designed around the workflow of the Linux kernel originally. That's why there are commands like "git format-patch" and "git am" which have help messages like "Prepare patches for e-mail submission" and "Apply a series of patches from a mailbox". It was also not that uncommon to see code patches printed in magazines or snail-mail letters.

tyingq 9y ago
This was my experience as well. In fact, source code control tools weren't that high on the list of success factors. Communication and distribution were more important. "Pull requests" were often just a tarball of diffs against a specific version, sent via email, and discussed on usenet. The code owner had to deal with the merge issues if the tarball wasn't built against "trunk".

mdekkers 9y ago

sourceforge

protomyth 9y ago

OpenBSD with anoncvs started in the 90's with showing a live cvs to the world. That was a nice change from release dumps.

user5994461 9y ago

It's very simple: "Just use sourceforge". https://sourceforge.net/ https://sourceforge.net/ GitHub didn't invent anything. They only made a better UI. SourceForge was more oriented toward software users. It handles the distribution of binaries. Imagine a time when 90% of software sites would redirect you to sourceforge to download them. Distribution was first. The repo and the issue tracker were in a secondary tab. GitHub is oriented toward developers. It shows you the repo as the main page. You can't distribute binary releases (well, it was added recently). GitHub took over by attracting developers right at the start of the project. They need a repo, here it is. UI is nice, site works. The new projects eventually become well known and carry the reputation. SourceForge went downhill slowly with their shady practices. Ads, confusing pages and software repackaged with adware.

brians 9y ago
I remember being so confused by SF in its first years. Why would I want this, instead of CVS in AFS or over SSH? The idea that most software developers wouldn't have routable workstations, wouldn't have servers they controlled---that we wouldn't keep running networked file systems over the Internet---totally not obvious. The SF people foresaw a need in a changing world and got in front of it.
- pjc50 9y ago
  When and where were you that use of AFS was normal?
  - scolson 9y ago
    Not the commenter, but guessing AFS = andrew file system. This has been used for decades by a lot of the universities who were original internet consumers (MIT, Univ of Michigan, Carnegie Mellon, etc). In fact, it is still used by a lot of these universities and research institutions for data sharing and distributed computing, afaik.
    - danieldk 9y ago

hprotagonist 9y ago

It's worth noting that kernel development is still git + emailing patches around. It's too big for github/gitlab.

algesten 9y ago
What are you talking about?! https://github.com/torvalds/linux https://github.com/torvalds/linux Linus just don't agree with GitHub's pull-requests, so he doesn't accept such that way.
- quadrangle 9y ago
  You know about the kernel but not how to use your compose key‽ The fact that the kernel is mirrored on GitHub doesn't mean they use it in any sense like the way most projects use it. The real point is that the way Linux is developed has no relation to that GitHub mirror, so it's an example of open source development outside of GitHub.
- cyphar 9y ago
  I don't think anyone is claiming that GitHub can't host a mirror of a big git repo. That's not what parent is claiming is "too big". The problem is that the tooling for reviewing changes on GitHub doesn't scale and also doesn't match the model that Linux uses (deputies). Hell, I would argue large projects like Kubernetes and Rust are hitting these same issues. Also, that is just a mirror. The real hosting is on https://git.kernel.org/ https://git.kernel.org/.
  - algesten 9y ago
    Ok. I read it as "too big" size wise. Of course I know it's a mirror.

holydude 9y ago

Many people commenting here say that github did not bring anything new to the game. While that is partially correct we know that bringing only new things does not matter. What matters is slick,seamless polishing of existing ideas and execution. There are many examples of just reinventing the wheel and making it more beatiful and usable. To your original question: sourceforge was huge, google code, cvs, bazillion of local "downloads.com" etc

pluma 9y ago
I think it's incredibly naive to underestimate how much value GitHub provided by shifting the focus from users to developers and putting contribution front and center. SourceForge was primarily a way to distribute software. While most software would be under some form of open source license you'd often also find plenty of projects that either weren't really open source or hosted their source code elsewhere. And when the source was on SF, the web interface for browsing it was unintuitive and unhelpful. The most intuitive use of SF for non-maintainers was downloading releases and leaving comments. There was a separate space for filing issues but it was pretty much just another forum. And let's not forget that contributing code also meant you had to have experience with using SVN (or CVS, I think?) because learning that at the time felt even more daunting than git does now. Before THAT you were mostly confined to sending patches to newsgroups or mailing lists as far as I recall. I open sourced a single project on SF and beyond that never really got into open source until GitHub made it painless and straightforward -- I doubt I'm the only one.
- k__ 9y ago
  So true. Coming from times when I had to download my PHP libs from SF and trying to get some fixes landed, everything's much different and better now. I get everything via a simple npm install and landing PRs has never been easier.

ing33k 9y ago

Others have pointed at sourceforge , google code . There was also Freshmeat ( renamed to Freecode ) https://en.wikipedia.org/wiki/Freecode https://en.wikipedia.org/wiki/Freecode

roadbeats 9y ago

See suckless.org for a good example. Personally, I used to put them under my personal website, seperating each to a subdomain. If you'd like to see source code, you'd download it.

pjc50 9y ago

Kids these days :) Richard M Stallman will tell you of the days when people mailed tapes back and forth, but that was in the 70s. A key development was Larry Wall writing "patch" in the early 80s, which he posted to USENET. That made it much easier to start mailing changesets to the maintainers of programs. This workflow continued basically until Linux wrote "git" in the late 2000s. The idea of the pull request formalises the process of emailing Linus. In the late 90s and early 2000s, most projects had their own CVS server and mailing list. Mozilla decided they would write Bugzilla to keep track of bugs. Sourceforge grew up as an ad-supported hosting and distribution server for CVS (and later SVN) repositories.

hk__2 9y ago
> until Linux wrote "git" in the late 2000s typo: it’s "Linus".
anjbe 9y ago
Worth mentioning also that in the early through mid 90s, projects hosted their own version control servers, but you needed an account on the development machine to actually have access to the repo. Most people only had access to release tarballs, not the full history, and couldn’t (e.g.) make diffs against CVS head. OpenBSD developed anonymous CVS, which gave anybody the ability to get the full project history even if they weren’t a developer. (That’s actually where the “Open” in OpenBSD comes from—it refers not to the license, but the development process.) Here’s the anoncvs paper: https://www.openbsd.org/papers/anoncvs-paper.pdf https://www.openbsd.org/papers/anoncvs-paper.pdf OpenBSD’s latest release song was actually about this topic: https://www.openbsd.org/lyrics.html#61 https://www.openbsd.org/lyrics.html#61
Turing_Machine 9y ago
"Richard M Stallman will tell you of the days when people mailed tapes back and forth" There's a story of an AT&T source tape being left on a street corner by an anonymous donor in the early days of (I think) BSD.
luckydude 9y ago
Larry Wall's patch was a huge step forward. In case you don't know what this is, it takes the output of diff and tries to apply those diffs to your source files. If diffed baseline and your baseline are the same then the patch will apply cleanly. If they are close the patch may apply cleanly (though it may have some messages like "offset 3 lines" which tells you that it found the place to patch 3 lines away from where it thought it was). There was a big period of time where the Linux kernel work was all diff/patch based. So the answer was (a) a mailing list that everyone was one, (b) some way of knowing a baseline (you could just have a baseline tree and a copy of that and you diff -Nur the two trees), (c) email the patch. All that worked but it didn't scale which is why Linus switched to BitKeeper and then later Git.

catwell 9y ago

I started being involved with Open Source in the early 00s. For source code, people typically used CVS or (later) SVN. The repos were hosted by platforms like SourceForge, berliOS or Savannah, or on the projects' own servers. Releases were tarballs hosted on servers. End users sometimes got software by downloading and compiling those, but there were already package managers and repositories. Before most people had Internet at home, at least in France, you could buy CDs with Linux packages at the press store. Communication between developers typically happened on IRC for casual chat and through mailing lists (hosted with mailman) for serious stuff. Sometimes there were web forums to communicate with users. And conferences, of course. A lot of projects were developed by people who were physically close, for instance people who met at the same Linux Users Group. Some projects used bugtrackers actively, some did not. It was mostly Bugzilla or Trac. That did not really change too much with GitHub. The main difference in my mind is that forking was usually regarded as something bad. You would not fork a project and then send pull requests to fix issues. For a simple fix, you would generate a patch and send it to the project's mailing list, where it would be discussed and eventually checked in by one of the owners. If you wanted to really engage with the project, you would become active in the community and eventually get commit rights. I think what led to that change of model is: cheap branching in Git, GitHub as a platform with users account, and maybe Continuous Integration making it easier to validate patches.

jedisct1 9y ago

Sunsite. And new releases were announced on Freshmeat.

kem 9y ago

God this more than anything before has made me feel old. Is GitHub that strongly associated with open source that alternate models for open source development are a mystery? Somehow that's really troubling to me in terms of the open source paradigm, even though I generally see GitHub as a very positive thing. I don't mean that as a criticism of the poster, just saying that it feels like maybe there's some room for more diversity and competition. I think others have given pretty good descriptions of pre-GitHub models, so I'll leave that to them.

geraldcombs 9y ago

Self-hosting. SourceForge was certainly popular at one time, but it didn't own the market, certainly not to the extent that GitHub currently does. Many projects were simply run on whatever hardware and connectivity the project could scrounge up. For a project I founded I initially bought a used SPARCstation on EBay and installed CVS, Apache, Mailman, Bugzilla, etc. I traded consulting time with various local ISPs in exchange for bandwidth.

herval 9y ago

SourceForge. Before that, webpages in a "ring". Before that, Usenet/BBS forums. SourceForge, specifically, is basically the same as GitHub

tzs 9y ago

In addition to the repository hosts already mentioned, there were also some specialized services for open source for a particular type or class of use. A good example was CurseForge.com. They provided free SVN, Git, and Mercurial hosting for add-ons for several popular games, along with per add-on issue tracking and discussion. Add-on authors could host at other, general purpose, hosting sites, but there was an advantage to using CurseForge. Curse.com had an application called the Curse Client for end users. The Curse Client was an add-on manager. It integrated with CurseForge, allowing you to handle searching for, downloading, installing, and updating add-ons from CurseForge right in the Curse Client GUI. Curse and CurseForge are still around, and support a variety of games. I'm not sure what the deal is with repository hosting there now. I saw some documentation that suggested you are supposed to use an external repository and provide a link to it as part of your CurseForge project settings, but I also see a lot of projects whose repository is on curseforge.com.

chrisgoman 9y ago

Self-hosted Trac (without the "k"). It was usually trac.domain.com

payne92 9y ago

zipped source tarballs w/ anonymous FTP downloads. Then you emailed diff patches to maintainers and prayed.

NoCanDo 9y ago

Sourceforge or SVNs. Github didn't revolutionize the open source landscape. They just gave everybody an easier-to-use GUI.

skyzyx 9y ago

I managed SimplePie using a self-hosted SVN repository for the first 5 years of its life. Contributions exploded AFTER moving to GitHub. The AWS SDK for PHP started life as a project called Tarzan on Google Code. GitHub made code feel more accessible. It made contributions easier. It improved the visibility inside the project.

oceanghost 9y ago

Hi, I was one of the original contributors to Gentoo in the late 90s (back then it was called Enoch). We had a private FTP site we stored source code and packages at, and we all downloaded regular snapshots of the system. I would archive milestones to CD-R out of a sense of duty+fear. It sounds primitive, but at the time source control was something only wealthy corporations had.