Slashdot Mirror
Tim Perdue on GForge & Building SourceForge
Steve Mallett writes "I've just posted an interview I did with Tim Perdue, former co-'head honcho' responsible for developing SourceForge. You'll either love it or hate the interview, but it's on his new project GForge, a fork of the previously open source code running SF, while he shares some insight in what seems like a miracle that SourceForge was built at all." Obviously Slashdot's parent runs
SourceForge, so insert whatever mental disclaimer and conspiracy theory you
want here.
Tim Perdue was one of the founding architects of SourceForge, the open source project management and website, hosting thousands of projects and home to over 500,000 developers. Tim is also known for having built both GeoCrawler and PHPBuilder. OSDir asks Tim about his days at SourceForge, what happened behind the scenes, and his latest project, GForge, a scaled down and enhanced version of Alexandria, the code that VA closed to sell as proprietary.
OSDir.com: The last time I'd seen your name pop up, previous to your new project, was as a head honcho at SourceForge.
Tim: Well, I was only "head honcho" briefly after Tony Guntharp had moved to engineering and before Pat McGovern was brought in as manager. The funny part was, no one told me I was "in charge", until I was doing a presentation to the board of directors and I was introduced as the manager.
Other than those few months, I was lead developer, which also didn't mean a whole lot, since most of the time there were no other developers, and certainly no road map or plan for how SourceForge would progress. I essentially just built out and developed things on an ad-hoc basis as I saw fit. Drew Streib was the other founding developer for the first 9 months or so and he did the same, and we loosely coordinated our activities and criticized each other's code.
OSDir.com: What was it like at the helm of what is one of the biggest open source projects in existence?
"There was no budget - the hardware was literally repo's and rejects from tech support. This was a major irritant later on, when you would hear at company meetings how such-and-such a manager had provided all the budget and support to make this happen."
Tim: It was certainly exciting to see how outrageously successful it had become, and to be in a small group that was building it out. It was all very ad hoc, very loose. It was a lot of fun for the first several months.
After that, there started to be some tension. Drew and Tony had very serious personal conflicts. We all had pretty strong personalities, bull-headed so to speak, and that raised tensions. At the same time as exponential growth was occurring, we didn't have support in terms of staff or budget. Actually we had less staff with the departure of Drew and Tony a few months after launch. So it was Uriah and I trying to keep this monster growing totally by ourselves. I guess Quentin had joined to handle support requests at that time. But it got really, really ugly for Uriah, as he tried to admin and keep all the shoddy hardware running 24x7 by himself. That was really shameful. We sent some seriously nasty messages to upper management, which went completely unanswered and ignored.
At the same time, most every manager in the company was trying to take credit and grab control. At company meetings you would hear how the whole thing was the idea of so-and-so and they supported it from the start. The reality of how SF started was very different, and the official vision for SF was also very different than what we wound up with. Probably only the founding team remembers know how SF really started and what the original management vision was.
OSDir.com: What was that original vision?
Tim: The real goal of SourceForge was to build a site aimed at IT managers. Believe it or not, that was the real goal. Nobody working there, except Uriah and Larry Augustin, knows that. Everyone else has been fired or quit.
"My main interest is turning this into a client that you can detach and take with you."
Around August of 1999, VA had received back a report on "name recognition" from dataquest or something, and no one knew who VA Linux was, and so no one would buy VA hardware. So SourceForge was going to be a site where every IT manager in the world would come and get information, statistics, surveys, ratings, all that sort of stuff, on open source software and see the VA logo. The hosting of projects was sort of an underpinning to that.
Anyway, the management wanted this whole thing built out by Comdex in November - 2 or 3 months away at that point. There was no way that could happen, so Tony went to bat with management and they agreed to split it into 2 phases. Phase I was project hosting, which we could slap together by Comdex, and Phase II was the IT manager stuff, by February LinuxWorld in 2000.
So the original gang of four was shut in a small private office in SunnyVale and we just started whipping up this system. There was no plan. We just winged it. If we thought of something cool, we slapped it together and that was it. There was no budget - the hardware was literally repo's and rejects from tech support. This was a major irritant later on, when you would hear at company meetings how such-and-such a manager had provided all the budget and support to make this happen.
Anyway, Phase I was such a hit that management never again mentioned Phase II after the Comdex launch.
Shortly after the launch, you started hearing about "Server51" from Andover, which we viewed as a direct ripoff of SourceForge. To counterattack that, we were told by management to release the SourceCode right away, even though it was obviously crap that was hard-coded to run on SF.net.
It was at this time that VA's latest strategy was to distance themselves from Linux - the lustre was fading on that buzzword. Now VA was an Open Source company. They gave MySQL a couple million to GPL their code, we gave away our code, and so did Slashdot. I really expected that the company was going to change its name to something around "open source", which was quickly becoming a better buzzword than Linux. It was really VA management that started the open sourcing of the code.
"...why shut off a valuable open source tool, when you aren't going to wind up with a viable business. And today, SourceForge Onsite has less revenue than your average gas station."
The concept of shrink-wrapped SourceForge software didn't come about until someone from IBM or HP casually mentioned that it would be cool to bring SourceForge inside their company to manage projects.
OSDir.com: How did you part (from SF)? Was it on good terms?
Tim: A little bit of background. As the rest of the company imploded, we had a lot of those people come over and all of a sudden they cast themselves as the "experts" on everything from building a great development platform to running the development process. And of course, I didn't know shit from shinola, which I always found rather amazing since I built much of what they were fighting over.
What ultimately caused me to leave was getting called on the carpet about a post I made in the alexandria-developer discussion forum. A lot of people were wondering why the CVS tree was unavailable, when was the tarball getting released, etc. And in response, one of the new managers posted a response to the forum which was... less than frank. Mind you, it was my name attached to the happenings around the SF code, and I think everyone in this community of users knew me by name, so I did not appreciate the handling of this. It would have been far better to simply be frank and open with what was going on.
Internally, we all knew the code was closing and the entire company was being reorganized around the proprietary code. Practically everyone I knew had just been fired, and what was left was all the managers that were desperately clawing for a seat on SourceForge as everything else just disappeared around them. I'd say this was far bleaker than a year earlier when Uriah and I were desperately trying to hold SourceForge together with duct tape and bailing wire, and I had had enough.
Anyway, they weren't able to sell any of the paid installations while the code was out there, but the management was maintaining a facade that the code was coming,and that's what they said on the message board. It was a lie. I found that intolerable and I posted a list of places that people could obtain the latest code (various forks on the internet, like debian-sf, sfportable, and sf-genericinst). Well, I got yelled at and I literally said 'Farewell' right there on the spot. The manager in question was headed out for a weeklong vacation and told me to think about it while he was gone, but I decided a few hours later. And that was it.
OSDir.com: How do you feel about the closing of the alexandria code base by VA and subsequent projects that built on the still open source, but older, versions?
Tim: I had mixed feelings. I certainly understood that VA was "tiptoeing in the graveyard" as the CIO, Steve Westmoreland, told me. They were pretty desperate for a strategy after the hardware business imploded, and they were convinced they could build a business around SourceForge as a shrink-wrapped product.
But at the same time, I didn't think it was a good idea to close it, mainly because I didn't believe they were going to have any success selling the proprietary product anyway. So why shut off a valuable open source tool, when you aren't going to wind up with a viable business. And today, SourceForge Onsite has less revenue than your average gas station.
OSDir.com: So now the "valuable open source tool" is back in action in the form of GForge. Tell us what GForge is about. How much is it philosophicallyand practicality motivated?
Tim: I agree philosophically with the Debian-SF team that this is a fundamentally valuable tool for organizing and streamlining development, whether it's open source, or whether it's internally at companies. That's why I'm working closely with Debian-SF to ensure that we share a common codebase as things move forward. If developers can crank out more and better code based on what we're doing, that's something we can be proud of.
At the time that I had left VA Linux, I was aware of downloaded/free copies of SouceForge running at the biggest of the big companies, from GE, Sony, IBM, Xerox, and Intel to others I can't remember anymore. IBM DeveloperWorks even runs part of their main website on old SF code, and so does the Free Software Foundation. So the open source code is having a huge impact. Someone really needed to pick it up, and after a year of trying to get others to take up the slack, I finally did it myself.
OSDir.com: Have you had people approach you at all to work on a new version of the SourceForge code?
Tim: Yes, and interestingly enough, many of them are old haunts of VA.
OSDir.com: Have they approached you about building a new SourceForge-like site?
Tim: Well, I'm not interested in putting up a generic project hosting site myself, but there's some serious interest out there in either doing that, or in funding a rewrite of the entire codebase so it is distributed. I've put some thought into how you could restructure this whole thing so it is sort of like gnutella or freenet.
That's a big rewrite, but I do know there are a couple people willing to throw some cash at that and more people willing to code it.
OSDir.com: What kind of improvements have you incorporated into GForge?
Tim: There's not a whole lot in GForge 3 that wasn't in 2.61, aside from a lot of cleaning up and removing hacks that made it harder to install. Part of the "Open Source Strategy" for a while, was that we would make SF really hard to install, to encourage people to buy the commercial product. As a result, it was a real ***** to install.
"A single machine is sufficient. There's not much load involved for a small number of people. You could probably run 1,000 or more people on a single machine given how scalable this thing is."
OSDir.com: Ouch. Well it certainly was, ahem, difficult to install. The previous version of OSDir.com used the code. Notice it's no longer the case.
Tim: The most obvious change to the user is going to be the tabbed theme. This is something I kept trying to do with the official SourceForge, but could never get buy-in on it. Jabber support is almost complete, foundries have been removed, and as I mentioned, it's a lot easier to install. Tons of extra code which was needed for the ultra-scalability of SF.net, is no longer necessary.
OSDir.com: Ah classic. Is this the "Perfection is achieved now when there is nothing more to add, but rather when there is nothing left to take away." thing?
Tim: Hehehe, well it is kind of funny that my first instinct was to tear out things that irritated me for a long time, like the foundries. But there is a lot that can/should be added, and a lot that still needs to be rewritten to make it "cleaner" and more maintainable.
OSDir.com: What kind of features do you see being added that may have been hampered previously by other concerns?
Tim: My main interest is turning this into a client that you can detach and take with you.
OSDir.com: How does that work?
Tim: Think of it more like CVS for bugs, tasks, etc. You can copy it to a local machine and check it back in if you made any changes.
This was something Drew and I first talked about 1-2 months after the original SF launch, and a lot of groundwork was laid in the codebase to do it. But it was never realistic given the difficulty in keeping the site running from day-to-day. There were also concerns at that time that you wouldn't be able to show any ad banners, etc, so you really wanted them to visit the web site.
GForge also needs a "real" project manager - the current task manager is pretty weak. The forthcoming project manager will be based on the tracker, sort of a superset of tracker functions with calendaring and resource management added in for kicks. So when a bug is submitted, or a patch, you can easily spin that into a task, and join lots of bugs together into a task in the project manager. It needs gantt and pert charting too.
OSDir.com: Do you think you'll be able to muster the same kind of developer support and enthusiasm for this product as there was for developing the code under the SourceForge brand? Realistically, would that be an enviable thing?
Tim: There is some hope that by having this out there, it applies pressure to VA to open their codebase again. If so, then everyone wins. Aside from that, I can see there is enough interest already to make good use of the code, and I have already received a ton of code back from the debian-sf crew.
OSDir.com: What has the debian-sf crew been able to contribute?
Tim: They have replaced most of the strings on the site using the internationalization support that we built into the code, but never really used. So they've got complete translations in English, French, Spanish and Korean. And of course, a really slick and easy installation process.
OSDir.com: You've indicated that GForge is at Version 3.0 Is it ready to roll for anyone who wanted to set it up right now and take on some projects?
Tim: Yes, although we're currently on the pre4 release. Pre5 will complete the jabber support in a couple of days, and the final release in a couple of weeks will roll in all of the debian-sf changes.
OSDir.com: How much hardware in involved in using GForge today for say 1-25 projects?
Tim: A single machine is sufficient. There's not much load involved for a small number of people. You could probably run 1,000 or more people on a single machine given how scalable this thing is.
OSDir.com: How many folks have deployed GForge to date?
Tim: I don't know that, but there have been about 1,000 downloads so far, and I've heard from a lot of people who have installed it successfully and were happy with how easy and fast it was to install compared to old versions.
OSDir.com: What kind of feedback are you hearing from your adoring fans regarding GForge?
Tim: People are excited. I'm particularly surprised how much energy there is among the ex-VA employee crowd. Tony Guntharp, who was the original manager of SF, is offering to pitch in some interesting stuff he has written. Michael Jennings is working on an RPM installer. We've got an internal corporate user offering to pitch in ClearCase and Synchronicity support. There's a lot going on. I'm pretty impressed.
OSDir.com: Is there anything else, Tim, that you'd like to add regarding GForge?
Tim: Only that if you have time and energy, we could use your help rewriting portions of the codebase. There are a handful of old, crusty sections which have never been "upgraded" over time, and that would be a huge help.
Other than that, if you need to manage multiple development projects, try out the code and see if it helps.
OSDir.com: One last question. What does the "G" in Gforge stand for?
Tim: GNU I suppose.
It sounds interesting. My company looked into sourceforge briefly but the deal is that it was so big and so costly that we simply as a small auxillary division of a larger company could not afford it for a software project needs.
I am looking forward to trying this out.
ACK
actually Tim specifically took out a lot of the hackery that allows SF.net to scale to such ridiculously huge numbers, which are not generally needed for small installations, which is what GForge is targeted to ...
09
Don't get me wrong ! this is not a flambait ! SF is a very nice service for the community and I sure thank VA very much for offering it. But ! because Tim Perdue was a PHP programmer they have decided built everything from scratch and they did badly. A new case of reinventing the wheel. I mean why did they build a new and very weak bugs tracker ? while Bugzilla even at that time was already doing a very nice job, they have built new forums very clumsy to use, I can go on and on ! they have build a new mail archive which works very badly while pipermail does a very nice jobs. Best of all they closed the source of SF and have completely stopped adding new features, they won't install the new version 3.1 which is now commercial only. Plus they don't answer bugs requests neither feature requests anymore. It's not google searchable, and It's really difficult to search in all the forums of a project in the same time as you must visit _every_ to do your search, I can go on and on and on.
The interview link is down right now, so please forgive me if this is already discussed there.
There is a similar effort underway called hovercraft which is the code that drives mozdev.org. The main difference i see here is hovercraft's focus is on using well established existing open souce tools like bugzilla.
http://hovercraft.mozdev.org/
I didn't "invent" sourceforge. The original SourceForge site was the efforts of:
Myself, Tim Perdue, Drew Streib, Uriah Welcome
We also got tons of help from the following people:
Quentin Cregan, Steve Westmoreland, Chris DiBona, Joe Arruda, Jeff Ritter, Patrick Wong, Neil Doane, Robert Flemming and Kit Cosper.
legit beef? oh heck yeah....
... sf.net was founded and the code was released ... for a while the OS repository was maintained and SF grew to be big and strong ... then VA decided to try to make money from SF (no problem here, yet) ... they started selling it commercially to a few large companies who found it very useful ... as time went on they allowed their techs to interact with the OS community less and less with regard to SF bugs, features, etc (tim can speak more on this than i can, as he was one of these techs) ... they also weren't allowed to do things like write any installation routines or clean things up to make it more portable ... then VA decided to completely close the (GPL'd) code to SF, and proceeded to remove all traces of it from the site ... the CVS repository was deleted, downloadable packages were removed, and forums were even moved to a new slightly-hidden project (and if you ask me, they seemed to get a lot of strange off-topic posts suddenly, but that could just be paranoia on my part) ...
... Roland Mas, of the Debian-SF project, had already been working on a SF2.5-based package, which is now included in Debian's Stable distribution (woody). ... it's a slick, clean installation, quite impressive, and a 2.6-based version is in the works ... it is expected that GForge and Debian-SF will merge at some point, too ...
...
basically, here's the deal
but, quite a few copies of SF2.6.1 (the last tagged CVS release) made it out the door before things imploded, and some folks took it up and ran with it
VA promised to release SF "2.7" in August of 2002, which has come and gone with not so much as a peep from VA
tim (or roland, or anyone really), kindly correct me if i'm off on these points, but this is how i understand the situation
09
Actually, Sourceforge originally came into being when Larry Augustin approached me about building what he called: ColdStorage. This was to be a huge code repository that would never die and make software always available. I had my hands full with the linux.com team I was building, and scrambled to find someone capable to build this beast. I lucked upon Tony Guntharp at the Linux Expo in NC, and the rest is pretty much history. Just an FYI.
Founder: Themes.org, Linux.com
http://theinterference.com/ || My band
I appreciate Chris' response to the Server51 ripoff comment, but I have to admit it's sort of half-true in that it was Andover's response to VA's SourceForge. In late 1999, Scoop and I and few others in Andover were pointed directly at SourceForge and we were told we had to respond to this (meaning we had to build something to compete with SF). So we started building Server51, it was certainly meant to compete with SourceForge (our concern at the time was that SourceForge could someday overtake or at least deflate Freshmeat as a relevant open source destination) but we honestly wanted to make something far better than SF was at the time, something more useful for devlopers and users, and make it kind of fun with the campy alien conspiracy theme (the idea was the server existed in an unknown location and everyone working on the server was part of some larger X-Files plot) . We didn't use any SF code (although it was available to the public) or SF architecture, we started from scratch and had our own ideas on how to make it all work, be reliable, fast, and scalable. I think what Tim meant by "direct ripoff" was "competetion". Well anyway after VA acquired Andover the intention was that SF and Server51 would be merged, but really Server51 was in beta and SF was already quite large so it was like merging and motorcycle with a bus... it didn't quite happen, just as well SF eventually improved anyway and I think its better that open source community has cooperation between Freshmeat and SF instead of having to announce new releases in two places at once.
I was close to the action from practically the beginning until VA and I parted ways last August. Those 4 are still some of the most amazing, versatile, talented people I ever encountered in the workplace.
BTW, Tony, you forgot to mention Usman Farman (ibr0w) and Geoff Herteg,
zeruch