Interview with Ben Collins-Sussman

Posted 10 Aug 2004 at 12:38 UTC by shlomif Share This

This is an interview of Ben Collins-Sussman (of Subversion fame) conducted by Shlomi Fish. It was conducted over the IRC at 23 February, 2004, and then over E-mail at 2 August, 2004. Final editions were done by both parties.

(This interview can also be found on Shlomi Fish' homepage

S.F.: Shall we start with the interview?

Ben C-S: Sure.

S.F.: Tell us a bit about yourself.

Ben C-S: Like what? I'm six feet tall, black hair, brown eyes. :-)

S.F.: Who you are? What did you do? Where have you studied and what? And that's 1.80 meters tall.

Ben C-S: Ah, ok. Um, let's see.

I'm 31 years old, and have lived in Chicago all my life. I have a B.Sc. in Mathematics from the University of Chicago, and have been working in computers for about 10 years.

By day, I work on Subversion... Collabnet hired me to help design and implement it in 2000. By night, I work as a composer. My collaborator and I write musicals and scores for theater around the country.

S.F.: Have you been a computer enthusiast during your teens?

Ben C-S: I did BASIC programming as a kid, between the ages of 9 and 12, then lost interest in computers completely.

At the age of 17, getting ready for college, I started getting into fractals I was generating Julia Sets on my apple IIe, and it would take 10 hours for each.

S.F.: <giggle />

Ben C-S: My calculus teacher in high school said, "hey, lemme introduce you to this guy, he can help you." And that's how I met my friend Matt Braithwaite. He had a 286 and Turbo C. I learned C by watching him implement Mandelbrot sets.

S.F.: Cool.

Ben C-S: Then I went off to college, and continued to write little toy programs in C for fun. My summers in college, I worked for the US Dept of Energy as an assistant to a professor who was simultating global climate changes on massively parallel supercomputers. That's how I learned Unix.

I remember in the summer of 1991, Karl Fogel and I were temporary roommates. (Karl and I have been friends since I was 13)

We both decided to install linux 0.95c on our 386 computers. :-) It barely worked. X windows blew out his monitor.

Anyway, after college, I spent many years as a unix sysadmin. But I finally decided that coding was more fun. I also got married. :-)

What else would you like to know?

S.F.: OK, that's fine for now. Now let's move to some other questions.

You maintain a homepage with various odds and ends. How much time did you spend maintaining it? How do you maintain it?

Ben C-S: I update the homepage about once a year, heh. My webpage is actually an svn working copy. My repository has a post-commit hook that updates the live working copy. It exists mainly for my relatives, not my friends.

S.F.: OK.

Linus Torvalds once said that 95% of programmers consider themselves in the top 5%. Do you consider yourself a good programmer? Do you think your programming skills have improved in time?

Ben C-S: Ha. My programming skills aren't very great, no. Honestly.

My strengths lie in organizing things. Whenever there's a hugely complicated topic that needs to be sorted out, everyone in the Collabnet Chicago office asks me to do it. For example, "go learn WebDAV/DeltaV" or "figure out what' s happening in that complex email thread; summarize all the positions" My talent is in being able to present high-level overviews of complex topics.

And I guess I'm also sort of the unofficial publicity/marketing guy on the svn lists. I seem to be worried about svn's image more than I should be. :-)

Really, when I code, it's not bad code, but it's not particularly clever either. But sure, everyone gets better over time. I guess that's all I can say on this question.

S.F.: OK. Do you ever look at your old code, and say: "who's the idiot that wrote this junk?".

Ben C-S: Only my perl code. :-)

No, seriously ,that has happened to me. If I look at the SVN code I wrote in 2000, there's a noticeable difference. That proves that there has been improvement.

S.F.: How much do you think social skills are important for someone who runs a free software project?

Ben C-S: I think they're everything. I think the reason svn is successful as a project is because of Karl Fogel's ability to be a good leader. I can point to other projects that have excellent technology, but failed to "coelesce" as a community, because of bad leadership.

Having good technology doesn't mean you automatically get a community. You need to socially manage people, not just review code patches. In fact, Karl is going to be writing a book about this very topic. :-)

S.F.: I'll look forward to it.

Hackers are well known for being socially-challenged and many times even acknowledge this fact about themselves. Do you think a hacker can have some rules of thumb to help him attract contributors and users for his project?

Ben C-S: 1. Don't be a jerk.

2. don't be a control freak. I belong to once open-source project where there's *one* guy who is writing the next-generation 2.0 version of the software from scratch, and he won't show the code to anyone. He refuses to put it into CVS until it's "done", because he doesn't want anyone to comment on it.

I guess that leads to

3. You are Not Your Code. Be ready to take criticism and grow from it. If you criticize others, do it tactfully, respectfully, and only give constructive criticism.

Be willing to toss your efforts and try again. Don't take anything personally. I've seen so many flame wars break out over this.

4. Set some social standards and stick to them. The svn project is an example of this: we expect patch submissions in a specific format, we expect people not to file issues without discussion first, we have a voting system.

I guess it's really hard to separate politics and social behaviors when email is your main medium of cohesion. But project leaders should lay down political laws (i.e "how does one become a committer? how does one vote?"), as well as standards of social conduct, and not be afraid to boot people out of the community if they ignore either set of standards.

Eventually. you end up with a community that polices itself automatically. You've selected the community for a certain type of "plays well with others" kind of person, and so the group learns to automatically reject the opposite sort of person.

S.F.: Now let's move to some questions about Subversion.

How did you get involved in Subversion?

Ben C-S: I was hired to work on it in 2000.

S.F.: Were you involved in it beforehand?

Ben C-S: No, it didn't exist beforehand. :-) Karl Fogel, Jim Blandy and I wrote the first design document.

S.F.: What do you like the most about working for it?

Ben C-S: I like the fact that it's Free. If my company were taken over by another, or suddenly collapsed, the software would still have a life of its own.

I can't imagine pouring sweat and blood into something, and then watching it die simply because a company, or creditors, say so. The lifetime of a piece of software should depend on whether or not a community continues to exist around it, not if a lawyer says so.

That's all. :-)

S.F.: You're getting paid to work on Subversion. How is life in the Subversion office?

Ben C-S: It's a great office. When we're not working on SVN, we play music on breaks. Each of us has an acoustic guitar by his desk. There's great value in being in the same room with someone.

S.F.: Which of the core Subversion hackers do you admire and for what?

Ben C-S: What is this, a performance review? :-) I admire Jim Blandy, for designing the repository. That was the single most important design in all of subversion.

S.F.: Long question now. I noticed a certain trend in OSS. If we take Perl and Python for example, they are two languages who compete for the same niche, yet are very different and incompatible. Now, we can hear a lot of Perl-bashing from Python people ("Here's some working 'ugly' Perl code - let's rewrite it in Python", "Perl does not scale to large codebases.", "All Perl code is ugly", "Perl code is not different than line noise" "Oh, I tweaked some Perl code - I feel violated".) etc.

On the other hand, criticism of Python among Perl hackers is much less covert and gentle. Most Perl people will be happy that an unhappy Perl hacker became a happy Python hacker. They hardly ever claim that Python is bad, just that they may not like it much, or that Perl works for them.

It seems that there's a similar situation with Subversion and Arch. Subversion users just confess that they like subversion and not that it's superior to Arch or to anything. Arch people, OTOH, attack Subversion all the time on various grounds.

It does not seem to be applicable everywhere. I didn't notice it much with Linux and FreeBSD.

What is your opinion on the Perl vs. Python, Subversion vs. Arch etc. phenomenon? Is it desirable? Is it inevitable? Were you guilty of it yourself? ;-)

Ben C-S: Hoo. Here's my analysis: basic human nature.

If there are 2 products to fill a niche, and one of them is more popular, the less popular one will always being trying to "tear down" the other.

The more popular product dosen't fling mud, because, well, it feels secure in its popularity.

But I suppose there's a cultural aspect too. In the case of svn vs. arch, we see a debate older than me. Is version control about managing trees, or about managing patches? Subversion chose a side, but doesn't believe it's the only side. Honestly, Tom Lord, great guy as he is, is completely convinced that patch-management is the only answer to the problem. And so his position spreads through the arch community. It becomes a sort of evangelism: "we must make everyone see the truth!" Whereas the svn guys don't think there's a single correct approach to version control. We have nothing to preach.

I guess that's all I have to say.

S.F.: OK. What is the most delicate part of Subversion and why?

Ben C-S: Delicate, as in, fragile?

S.F.: I suppose. delicate as in hard to tweak and get right.

Ben C-S: So you're talking, code-wise, not usage wise?

S.F.: Code-wise, yes.

Ben C-S: I would say for a user, the repository database is the most fragile thing. See the bottom of my anti-fud essay about that.

The hardest thing to tweak and get right (code wise) is the working-copy library. Versioning directories is very tricky. And doing everything via journaled logs is very tricky.

Everyone criticizes the libsvn_wc code about being a mess, but really, if we were to take the time to rewrite it from scratch, I don't think it would be any less complicated.

It just might have fewer bugs, due to hindsight. The problem is really complicated, and thus the code must be too, to some degree.

S.F.: Joel Spolsky said in his famous article "Things you Should Never Do - Part I" that one should never re-write a complete working codebase from scratch, no matter how "bad" it seems. In the sequel "Rub a dub dub" he demonstrated how to improve an existing codebase to increase its quality and maintainability. I can tell from experience that his recommendation works.

Subversion was a re-write from the grounds up done by many of the original CVS workers. Do you think it could have been faster to replace CVS (or CVSNT) component by component, thus yielding Subversion?

Ben C-S: CVS has no "components", that's the problem. If we had spent 4 years "improving" the CVS codebase, the result would have been a nice-to-read CVS codebase. Something very clean and intelligible. But that was never our goal.

It's not like we looked at CVS and said, "this code base is unmaintainable, let's start over". What we said was, "The quality of CVS code is irrelevant; there's no way we can add the features we want given CVS's design. We need a new design.". And when you're changing the fundamental design of something, no amount of "cleaning up" a codebase is useful. Cleaning doesn't change the underlying design. It just makes code more maintainable.

We had a specific design that was fundamentally incompatible with CVS's design. so Joel's approach wouldn't have helped us.

S.F.: It seems that the development groups of different version control systems each manage to come up with their own unique and interesting architecture for making it happen. What do you think of the architectures of the various version control systems that you've studied? Do you think the Subversion architecture is the best one for its purpose?

Ben C-S: I think that version control designs always involve tradeoffs. You push down one bump, another pops up. If you optimize the system around solving one problem, or making something easy, then something else becomes slightly harder.

What I said about Subversion and Arch earlier applies here. Arch makes it easy to exchange patches, but it's a little bit harder to produce a revision tree. Subversion makes it easy to walk any revision tree, but harder to produce patches.

Same with Subversion's order-1 branch/copy feature. It's nice that branching is so fast compared to CVS, but the tradeoff is that it's no longer easy to ask the question " which branches does this particular version of this file exist in?" So my answer is: there is no single best architecture, or best solution to the problem. Different groups or companies need to choose a system that "feel best" to them, and live with the particular tradeoffs.

S.F.: You happen to be a musician. Tell us about your Music experience, throughout the years.

Ben C-S: I started learning classical piano at the age of 6, but then switched to jazz piano and improvisation at the age of 12. I went to college and ended up playing piano in an improvisational comedy troupe, as well as conducting a number of college musicals. At the end of college, I met my friend/collaborator Andre, and we wrote our first musical theater piece. Since then, we've been working in the Chicago theater community for 10 years, and have written 5 more musicals (and done over 100 sound designs). Our latest musical is going to New York for a festival showcase, in hopes of future productions. See

I've also -- just in the last 2 years -- started studying bluegrass guitar and banjo. It's incredibly fun.

S.F.: What is your favourite band?

Ben C-S: Hmmmm, Led Zepplin, probably. Hard to say. I also love Ben Folds quite a bit too, as well as Indigo Girls, Phish, Radiohead.

S.F.: What is your favourite foreign (non English/American) band?

Ben C-S: I'm pretty musically ignorant. I don't listen to music in non-english languages. ;-)

S.F.: Name some of your favourite songs?

Ben C-S: Ummmm.... I'm not much of a music collector or listener. I spend more time producing music than consuming it, I think.

S.F.: How musically-eclectic are you?

Ben C-S: Banjo is pretty eclectic, no? I've started becoming a bluegrass fan just in the last year. It's a whole new world to me.

S.F.: You confess that your favourite UNIX flavour is FreeBSD. What do you like so much about it?

Ben C-S: Well, I used to use FreeBSD as my main desktop OS, but a couple years ago I switched back to Linux. I've decided that FreeBSD is the best server OS out there. But it's not as instantly easy to get going as a desktop workstation -- Linux is better at APM, cardbus support, thread support, usb HID support, wifi gui applets, instant CUPS support, and has tools like valgrind. All that stuff "just works" when you install Linux on a notebook; on FreeBSD, that stuff either doesn't work, or is tricky to get going. (Granted, this is all FreeBSD 4; I have not tried FreeBSD 5 yet.)

In my former life as a Unix sysadmin - I can say that FreeBSD, as an OS, is the "tightest" distribution out there. Linux distros feel like a bunch of pieces shoved together: a kernel, a toolchain, some user space apps, and so on. FreeBSD is one coherent system, everything compilable from source in a single 'make world'. It makes the system much easier to manage and administer... and the networking is incredibly solid. It's my first choice for a server OS, no doubt.

S.F.: Did you encounter an occasion where things did not work as they should there because the developer only tested it on Linux?

Ben C-S: Absolutely... most of the nice Gnome/KDE applets, for example, make assumptions about Linux device drivers.

S.F.: At the moment, you are being forced to use RedHat Linux, so the CollabNet core team's development environments will be more homogenous. (If I understood that correctly) Do you enjoy the experience?

Ben C-S: No, I'm not forced at all. I chose to use Redhat because it was a nicer desktop system, plain and simple.

S.F.: Do you think the fact that most Subversion hackers are UNIX-enthusiasts, is causing problems with making sure it is working perfectly with Win32?

Ben C-S: Not at all. I think that we have unbelievable win32 support compared to just about any open source project out there. Show me any open source project where the core developer group are all "win32 enthusiasts". It doesn't exist.

The real problem is simply a conflict of cultures. Unix culture promotes free software: the system comes with compilers and languages, and encourages users to tinker with everything. The line between "user" and "developer" is blurred. Windows is the opposite: it has "users" and "developers", and the latter category is a small minority of people off in the corner.

Subversion's team isn't the stereotypical "we're all unix-users, windows sucks, go away Microsoft" sort of group. We even have team members who work (or worked) for Microsoft. Many of us have win32 compilers, and a subset of us build and test on windows almost every day.

S.F.: Tell us a bit about your family. What do they do?

Ben C-S: Parents are psychologists; brother is studying to be an astrophysicist. A nice jewish family in the suburbs of Chicago.

S.F.: How did you meet your wife?

Ben C-S: At University, of course. :-)

S.F.: How does she feel about the fact that you're working on an open-source project for a living?

Ben C-S: She thinks it's nice, but doesn't care about computers at all. She just knows that it makes me happy.

S.F.: You seem to be interested in languages and linguistics (know German, Spanish, and started creating your own artificial language.). Can you tell us more about it? Do you learn any other languages? Do you wish to learn any others?

Ben C-S: I'm interested in relationships between languages (the biological "tree"), and studying how languages change over time. I'm always excited to learn new languages... but lack the time. Probably a separate conversation here...

S.F.: What was your single greatest surprising event in your professional life? ( Like something that someone did which you did not expect. )

Ben C-S: Finding a paid job opportunity to work on open source. :-)

S.F.: Name some of your favourite sites.

Ben C-S:

S.F.: What other open source projects have you contributed to besides Subversion?

Ben C-S: The only other one I've worked on is OpenRPG.

S.F.: Thank you for your time, Ben, and may you have good luck and success in all of your endeavours.

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!

Share this page