FOSDEM was as excellent this year as ever, but for me it was a bit bittersweet Seeing the other Classpath hackers is always a joy. This is the best programming community I’ve ever worked in, and so it is sad that OpenJDK is (slowly) eclipsing our efforts. On the other hand, as I’ve said before, the OpenJDK release is the best conclusion we could possibly have hoped for, and, I think, the most significant free software event of the last 5 years. (I said that at FOSDEM and Tom Marble, I think, asked me: “what about Open Solaris?”. I felt a little embarrassed, since I didn’t know what to say right then. But in hindsight, I suppose it would be that we already have an excellent, complete free operating system.)

I was very pleased to meet all the Sun guys at FOSDEM. I confess, I was a bit worried that this wouldn’t play out well, but in the end I think they did an excellent job of community engagement. The folks who came — Tom Marble, Mark Reinhold, Peter Ahe, Igor Nekrestyanov, Peter Kessler, David Herron, and Simon Phipps — are smart, interesting guys, not to mention fun to have a beer with. Also they listened (and continue to listen) to our concerns, they tried to understand our needs, and in general seem to get the spirit behind free software. Add to that a bit of praise for us and you have a model of how to initially transform a project from closed to open, no easy task.

Since I’m in limbo a little bit, job-wise, and since I don’t often see fellow developers, I took advantage of FOSDEM to ask folks what they think I ought to do. I think I was quite doggedly boring about it. Anyway, I did get a few interesting responses.

The first is that, by and large, people seem to enjoy what they do, or at least think that what they work on is important. Michael Meeks said I ought to work on OO.o. Miguel said I should work on Mono. Tom Marble thought that OpenJDK is the place to be. (Funnily, I don’t recall anybody thinking that incremental C++ compilation is interesting… though this fits the theory.)

Another approach I took was to ask people “what is the most important gap in free software?”. What is it we don’t do well, that we need, that nobody is working on? I ask myself this constantly — perhaps pointlessly, since coming up with an answer seems to be beyond me.

One nice answer I got was “documentation”, though of course I immediately shot that down

Andrew suggested that asking this question about free software was having too narrow a scope, and he pointed to OLPC as an example of widening of vision. And, this makes sense… as free software gradually takes over the industry, us old-timers do have to shed our old biases.

Overall I think the best (meta-) advice I got was from Miguel. When not pitching me mono hacking he said I ought to stop fixing bugs (super important!) and take a break to think about things. Good advice.

The big result is, I suppose, the obvious one. Well, maybe obvious to you, it took me quite a while to get here: nobody else is going to have the answer I’m looking for. Back to the drawing board, looking for something to be passionate about. It is a scary and difficult time.

Sunday afternoon was “DevJam”, a somewhat more interactive series of sessions designed as an information exchange concerning Java and packaging. Tom Fitzsimmons started this, with an excellent presentation. He covered all of our needs very well and remembered all kinds of important things that I’d forgotten.

Even though DevJam was probably the most important session, I lamely skipped out early. Jim Blandy was giving a presentation on gdb, and I wanted to see him speak and catch up with him in general; I’ve known Jim from his days with Cyclic (remember them?). Jim did a presentation on using gdb tracepoints in the kernel. He wrote all the code the week before FOSDEM, didn’t sleep that weekend, and was still debugging at 10am the day of his talk — now that’s a real man.

Jim said he’s looked at SystemTap a bit and liked it. He liked the idea of a real program having a type inference engine in it, and he thanked Frank Eigler and Graydon Hoare by name. (This seems out of place but it was in my notes and I thought it would be nice to pass along.)

On a related note, in case you care, gdb has been showing signs of life again.

Since I stayed at DevJam a bit too long I actually missed most of Jim’s talk. However, I’ve seen tracepoint demos in the past and they are very cool.

After this I stayed in the same room to watch Federico’s talk on profiling Gnome applications. Jim slept, I think :). Federico is an entertaining speaker and had some important things to say, but the talk was aimed at a somewhat more basic audience.

I’ve got one final FOSDEM post coming. It will be a bit more personal, I think, maybe touching on a few personalities, my future, and some other things.

Sunday morning I gave a talk on Fedora and Java, in the Fedora room. I met Chitlesh Goorah there, he was organizing the Fedora stuff at FOSDEM and did a great job. Also Paul Nasrat was there (and I saw him again later in Cambridge). My talk was sparsely attended, but as Andrew pointed out, 9am on a Sunday is not exactly a great time slot. I once again had a problem with the projector; sometimes my laptop works and sometimes it doesn’t. I know there is some obscure fix, but I never remember what; this problem is driving me toward a more minimal style, and this time I only had three slides.

Mark Reinhold started things off in the Classpath room. Mark is the chief engineer of the Java 2 platform — Sun sent a bunch of heavy hitters to this conference. Mark’s slides are available, so I won’t go too deeply into that. I liked how open Mark was about their processes and about wanting feedback. He mentioned the “spring” deadline as well, and walked us through which binary blobs will be required in the initial releases. (I noticed a couple nice things about Sun’s presentations overall: they all had little Creative Commons stickers on the title page — cute touch — and they used remote controls to change slides.)

He said Sun has about 200 engineers working on Java (geez) and another 200 on quality for the JDK. Sun has a very detailed, complicated, well-thought-out process flow for making changes to Java. They have various internal web tools for patch review and other things, and much of this seems to interact with their internal VC system. They’ve been doing distributed VC internally for 10 years, which is one reason they are picking Mercurial for their public version control. Why Mercurial and not one of the others? They did an evaluation, as did OpenSolaris. The OpenSolaris results are available on the web. I thought this long history of distributed VC use was interesting, since in much of the free software world this technology is still pretty new.

Next up was Igor Nekrestyanov, who works on the Java 2D team. He went over the various bits that will need to be replaced (I mentioned them and their potential replacements in an earlier entry).

I’ve been hoping all along that we could donate code from Classpath to help out here, but I don’t think the FSF can do this directly. So, we’d have to get permission from all the authors… and that is a pain. It looks like reimplementation is the only way to go

Peter Kessler talked about the VM and the GC. He’s worked on Java since before it was called Java; he added exceptions to the language. I didn’t seem to take many notes during his presentation, sorry Peter.

Then came Peter von der Ahé, who works on javac. Again I didn’t take many notes, but I do remember that he talked about having compiler APIs available in the future. He also mentioned access to the AST for Netbeans (good… this is one of the areas where Eclipse, IMNSHO, really has an edge). At the bar later he mentioned the “kitchen sink project”, the idea being to have a repository where people can try out their wacky language ideas by hacking the compiler.

Finally came David Herron, the Quality lead. The one thing I noted particularly is that they do “large application testing” at various points, this involves big applications, multiple machines, and a week’s worth of elapsed time. Sun’s efforts in this area make our efforts on gcj seem very small indeed…

I was a bit taken aback by Simon’s response to my previous post, I suppose since it never occurred to me that he’d actually be reading this. Anyway, to expand a bit more on the “positive focus” thing — in large part that is me writing advice to myself. Repetition helps me internalize important things like “ignore the trolls”.

Anyway, after the morning sessions we ate the usual lunch of extremely cheap sandwiches. I saw Rob Savoye there — Rob lives 30 miles or so from me, and we used to work together at Cygnus, but I only see him every couple of years or so, and always at a conference. He was a bit upset about something bad I wrote about Gnash (sorry!). Also he had somehow gotten his hands on OLPC demo hardware… nice. (BTW I heard that his blog about his Katrina relief work is really good, but I didn’t have a URL… if you have it, please post.)

Later we made our way through the crush to the Classspath developer room. I spend most FOSDEMs locked in here; Classpath is my project and aside from interest I also do it out of solidarity. As you can tell I’m a bit ambivalent about this on occasion; FOSDEM is a rich environment and there is no way to see more than a fraction of it — I always feel like I’m missing out. Also I usually miss the “donation return” meeting and I’m sure, just completely sure, that I should’ve won a nokia 770 this year

It turned out that the Fedora booth was next door to the Classpath room. I was pleased to see Tom (spot) Callaway there. Tom is a great guy. Also there was Max Spevack, who I met for two seconds. The Fedora guys did a great job organizing, they had a great-looking booth with nice artwork.

Mark led with a bit of an intro, and then the various VMs each had a presentation on their current status. Andrew Haley talked about gcj status, Dalibor about Kaffe (their big accomplishment for 2006 being “zero releases :-), Twisti about Cacao updates (many new ports there, also they hooked the binutils disassembler into hotspot to see what kind of code it generates), Rob on JamVM, and Ian Rogers talked about JikesRVM.

I think this was the first time a JikesRVM person came to FOSDEM. Ian entertained us with stories from academia — the counter-world to the one I usually live in — and also pointed out a funny idea of stuffing fields into holes in existing objects (I once looked at sorting fields by size to reduce object size, but this turns out not to save much). Maybe I’ll implement his idea in libgcj. Anyway, JikesRVM is cool technology and Ian had a lot of interesting things to say about why writing a VM in Java is a good idea. My notes say he also mentioned “parakeet”, which is an implementation of the parrot virtual machine on JikesRVM.

After this Mark did a more general status talk about what has happened in Classpath this year. The sheer volume of code is somewhat frightening.

That night we went to a vegetarian restaurant. Usually at FOSDEM we just end up at BXL, drinking beer and, in my case, eating a salad. This year I complained a bit and Petri did some research and with some help set things up. I felt a little bit guilty — I think I was the only vegetarian in the whole group — but it was wonderful. So, thanks to everyone who made this possible.

Some of the most important stuff at FOSDEM takes places at these dinners, or later in the bars. It is how we get to know enough about each to later interpret the tone of the email we exchange over the coming year. This year, Sun sent a whole platoon of JDK engineers on a charm offensive to win us over, which I think worked very well. It worked on me anyhow.

More on the various Sun guys in the next installment; on Sunday (resisting the lame joke) they took over the room…