Hacker News Reader

Comment by joeld42 | Hacker News Reader

I wonder how many of the HN readers don't remember how amazing sourceforge used to be. It was github before there was github. Easy, solid, project hosting for open source projects when that was often the main barrier to a good project getting more widely known.

I think it's one of the earlier examples of why "grow userbase now, figure out how to monetize it later" can be a bad idea. There was a time when (perhaps) a majority of open source projects were hosted on sourceforge, and I think they assumed that that would just translate to profit (or sustainability) eventually. When that didn't happen, and then they had to compete with Google Code, it began the slide down the slippery slope to the spam-infested site that it is today.

I hope that's changing. We need more options for hosting software projects, we can't rely on github being benevolent forever.

View on HN

To me what Github got right is making virtually no distinction between end users and developers.

SF tried to shelter "end users" from "dev stuff" yielding bad experiences for both. It's hard to use open source projects without getting at least a little bit technical. And it's hard to contribute when the dev side of things is kind of hidden somewhere. And yes, 2-3 clicks in the UI do make all the difference.

(I'm not replying specifically to this comment, just the whole discussion on what made GitHub better)

There is one thing about github I find annoying, which is inseparable from what makes it great. On SF, it was normal and the cultural norm to find compiled software along with the the source. I just spent my morning going back and forth with a dev on github because his project will not compile on my machine. I had found a cludgey hack to get working, but it broke any chance of future commits, and I figured he would rather have things working right. We both just gave up in the end, it wasn't worth the time for the estimated amount of effort I was going to put into the project.

Given the widely different environments people work in, it would be nice to see a bit more compiled code, and the concomitant dev effort before release to make sure it is actually possible for other people to use. On SF I downloaded so many little utilities similar to GNU tools that just worked and got me running in under a minute on Windows, and that is much harder to find today (except on SF--and I hope they come back to rock on!).

Thanks! SF will be rocking on soon enough.

Yep. I've been looking into places I can host compiled binaries on a budget of $0. GitHub Releases makes it difficult for security reasons and S3 doesn't fit in my budget.

Here's hoping for the SF revival.

Consider using our GitLab. We're working on making the build artifacts (binaries) more accessible from the tag/release page https://gitlab.com/gitlab-org/gitlab-ce/issues/13161

github doesn't really serve "end users" for the most part. github is designed for libraries and pieces of open source that will be used by other developers. So the "end users" github services are developers.

SourceForge is designed so that developers can serve their wares to end users. As in, someone with a Windows PC that wants to download a media player.

This difference in use case is one reason that SourceForge's top projects are all apps and github's top projects are all libraries/framerworks for use in development projects.

I have to disagree. One of my pet complaints about github is that my mother couldn't download anything useful from there.

SF had great downloads of binaries and I feel that's why it is still used by portableapps and others. If you can't be assed to click 2-3 times, you won't be the most valuable contributor anyway.

I don't remember sourceforge ever being amazing. From day one it was just CVS, with a terrible UI on top of it.

The only thing ever good to say about it was that it was free.

The UI wasn't great but how many reliable alternatives were there at the time?

I've never used SF to host something but if so many big open source project used them in the past they had to be doing more things right than just being free (that last one turned to be tricky in the end).

> The UI wasn't great but how many reliable alternatives were there at the time?

None, that's why it did so well :)

> The UI wasn't great but how many reliable alternatives were there at the time?

None. But when Github came around, there were also no reliable alternatives at the time (other than google code which sucked as well) and it was still a thousand times better than everything else.

Sourceforge was never "amazing". It was never "great". It used to be decent and the only player in the game. When Github came around, it quickly lost both those attributes.

I don't know why people fondly remember the "old sourceforge days" - I remember those days and they sucked.

Don't let the lack of competition forgive how bad something is.

I wasn't a coder in the Sourceforge glory days. But it was a great place for a frequent user of opensource software. I remember a lot of small projects only existence was their Sourceforge presence.

So Github might have provided a lot of new things but Sourceforge was really amazing at what it provided, when it provided.

Being the only player also means nobody gave small opensource projects any home when they did.

It wasn't really the only player; BerliOS and GNU Savannah appeared right around that time, and Launchpad isn't much younger. That said, they weren't really any better, in my opinion. At least SF had a bunch of mirrors, which was helpful back then.

SourceForge started out as a 20% time type project at VA Linux/Research. It was originally open source but went closed source when they pivoted from hardware to SourceForge. Savannah was a fork from before it was closed up.

They weren’t spectacular. But Sourceforge was, at one time, respected and relied upon by the indie Open Source developer community. Much time has since passed, they ended up sullying their own name, but this retraction, even though it might be a tad late, is welcome. (I was glad to see it).

Thank you. Trying to do the right thing here.

You probably don't remember that CVS was a wonderful improvement over RCS just like Subversion was a big step forward from CVS. Git is awesome but someday it will be outdated too.

The choice of version control system isn't especially relevant. A website like SF/GitHub is about building a surrounding infrastructure around the core VCS, not the VCS itself.

But SF never did that. It just took CVS, and added a thin web portal on top. Once you had the address of the CVS server, there was never any incentive to _stay_ on the SourceForge website for any longer than you had to. You'd just download the code and move on.

GitHub isn't like that. Code is treated as a first-class citizen, and the flow of the whole website is about letting people _explore_ code, and follow its changes over time.

Git may well become outdated, but that has nothing to do with the principles at work here. You could make "CvsHub" along the same "code-first" principles and it'd work great. A code hosting site needs to be about breaking down barriers between the code and its users, and that has nothing to do with CVS-vs-Git.

Sourceforge UI could be improved a lot but that doesn't mean the SCM has nothing to do with GitHub's success. They have thought it up alright but the GitHub model of network and connection is the extension of the natural flow of GitHub. And I'm sure implementing such a model of connection, forking and merging would be a whole lot more complicated to implement on top of CVS.

Sourceforge was a time when WebUI, webapps weren't that mainstream. Comparing that with GitHub when everyone wants thing to be much more web centric, when you a popular DVCS is ignoring what the past was like. Also it always felt like Sourceforge UI was primarily for the users of software and not the coder, which is very important for the more regular projects (not dev tools or libraries).

>The choice of version control system isn't especially relevant.

Actually it is, Distributed revision control has alot do to with GitHub's success as does the rest of their site. With out

CVS,Subversion, the other Centralized systems are just not as good by the nature for open source, peer-2-peer development models that SF was catering to.

To believe GitHub could have been just as successful it is was called SVNHub is to deny reality. GitHub owes most of its success to the underlying technology of Git.

Git is powerful, but it's certainly not the end of version control. Most importantly, it does centralized version control (which is what most of us are using in practice) very poorly.

features like binary file locks, large binaries, large repo /history and small working copy are either unsupported or only being added with extensions.

Setting up a multi-gig git repo for hundreds of users with binary assets that require full history but also require locking etc. is painful as hell.

Sourceforge added svn and git later and also CVS was advanced when it was released. Funny or ironical GitHub also provides centralized version control like svn/cvs via pull requests, which I am not a very big fan of due to unlimited, useless fork lying around without real purpose. So actually github undo what git was set out to achieve, provide a distributed version control.

As stated earlier with the closure of Google code the viable alternative for github are not many, so more the competition better it is. Sourceforge project of the month and list of end-user project pages is nicer then Github. Also I use ohloh.net to compare multiple-project and their activity even if they are on github. If sourceforge can make similar analysis, will use them more.

Is ohloh even called that any more? And do they even keep in sync with repositories? Last I checked the LibreOffice stats, they were way behind.

It is called openhub.net and yes they are bit behind in pulling repository data. But it still gives nice project comparison and also activity charts. If some projects can't be found there, personally use python with pandas and matplotlib to generate the same, but takes a bit of time.

I'm glad to hear that. It's a cool idea, but only works if they synchronise frequently.

The UI was fine by that time's standards. And it was not just CVS but also SVN, which was what everyone wanted to use (git didn't exist).

Free, supported and turnkey.

Pre-sourceforge, you had to hustle and find a host or pay for it, build a web site, setup a mailing list on it, build all the infrastructure from scratch. Back it up, maintain it, etc.. Bigger projects had an easier job of those things, they'd even have volunteers to do the work. It says nothing of smaller projects or how you started a new project.

Sf dramatically reduced the friction to get things going. Then they also supported different tooling, cvs, svn, others, they had build and test servers, if you want to see if your stuff builds right on alpha running Debian or something else you didn't have access to, they provided that. They invented free code hosting, made tons of mistakes trying to monetize it but at least the first few years their heart was in a good place.

Plurality is good with this stuff, I don't know if it will be good or bad, but github will be different in five years.

I agree SF was awesome in it's day. The Windows guys never really got on board which may be why some of the commentators here don't agree. It was an Open-Source thing before Open-Source was accepted by corporate America.

You can say the UI wasn't great but for it's time it was great. You have to remember this was way before any of the the Web tools people use today, except for PHP, but few use PHP now. These were the days when you had to support IE6 and every browser behaved somewhat differently.

I don't remember it being awesome, ever, just less crappy than the alternatives. It was like a bit of flotsam in an ocean, better than fending for yourself, marginally. You could roll your own hosting, but it was obnoxious and expensive. SourceForge was, at best, the Geocities of open-source hosting.

It really hasn't improved much since then.

Also a lot of people still use PHP. I really don't know where you get your numbers. It's just that there's no longer just PHP or Perl, there's dozens of languages that are in play, but the number of PHP devs is surely way higher than when SourceForge launched.

We will vastly improve the UI in time. Just need more than 2 weeks ;)

Tell me, were you mainly a Windows programer during the SF heyday?

I know people use PHP. It's not popular with the cool kids. Maybe you didn't get the memo.

No. I've was raised on IRIX, SunOS and Linux, and as much as I loathe everything about PHP I won't delude myself into thinking nobody uses it.

The industry is a lot more than just "the cool kids", you know.

I think it's important to separate its positive effects on the community from the merits of its UI. To me, SourceForge epitomized the "it's complicated for us, so it should be complicated for you" school of UI design. The owners didn't appear to do much to improve that over time, instead treating the site much like toxic code that no one dares to modify for fear of breaking something. No doubt that was a huge factor in their decline, which fed into the user-hostile cash grab tactics. However, the fact that it existed at all when no alternatives were available was great. I wish the new owners the best of luck in turning the site around.

Thanks. We will improve the UI significantly.

yes. and what about the Github after Github? If open code runs everything its going to a pretty important public utility.

Github has paying customers, I don't think they need to figure out monetization later.

Source Forge amazing? Tell me you are kidding!

Reply on news.ycombinator.com