ELS 2008 report #
hacking, May 27th 2008
Thank You, for everyone who worked to make The 1st European Lisp Symposium the roaring success it was.
For me, the symposium started by having dinner with other attendees (and some organizers) on Wednesday. This was obviously the right thing to do, because the whole conference went in the same vein: great people, good food, excellent wine. No presentations on Thurdsay -- but if you ask me, they are not really the thing that makes a conference tick for me.
Thirsday was the first day on the official program. After the initial welcome session, and kibbitzing over coffee people split into two tracks: the work-in-progress track, and birds-of-feather sessions. Not having submitted a WIP paper I did not attend that track, but from what I heard it was definitely worth attending.
- The BOF track started off with a session on concurrent and distributed processing by Sebastin Gonzlez. Questions of code mobility, identity, and structure of communication between nodes were talked about. Sebastin is working on a distributed and concurrent processing system, and has obviously spent quite some time thinking about the issues. From my perspective the project will be interesting to follow, and it is especially interesting to see what sort of things they will want from the underlying Lisp -- but generally speaking my own interests are on a slightly lower level.
- Next session was on image processing APIs, chaired by Matthieu Villeneuve -- working towards possibly consolidating Imago and ch-image into a single library. I admit I spent most of the session hacking on SBCL and checking my email, but I gather that those who participated properly found it engaging. Not quite sure were it ended up.
- Third was the CLIM (or perhaps rather McCLIM) session. I'm sorry to say it did not go as well as I would have hoped. The McCLIM hackers who were prepared for it had prepared to talk about McCLIM internals and future directions with each other, but due to the largely non-cognoscenti audience it turned into an impromptu CLIM tutorial/demonstration, which no-one was prepared for: demo-effects were plentiful, and direction was a bit lacking. Could have been worse, but could have been a lot better. It would be really great if someone could make a screencast to showcase the stuff that should have been in the demos -- but time being in a globally short supply I totally understand if that doesn't happen anytime soon.
- Lunch in a nearby resteraunt was simply excuisite -- and part of the symposium fee. Quite probably the mellowing effects of good food had a lot to do with the success of the following session... (Actually, maybe lunch was before the CLIM session? I forget.)
- Final, quite impromptu, session was on a future successor to ASDF. Andreas Fuchs, I, and several other intrested parties had realized this might be a good time to hash some things out face to face, and took the opportunity. Several conclusions were made -- in no particular order: Purely declarative system definitions are good: loading a system definition from a file should not use LOAD, just READ, and then some sort of an EVAL-SYSTEM. System definition extensions required for doing operations on a given system should live in a separate file. Manually managed fine grained dependencies are too brittle, and complete automated dependency management is too hard. Therefore encourage module-level dependencies (with stronger dependecy semantics then ASDF) as a half-way house between purely serial definitions and brittle file-level dependencies. Protocols are good: extending existing operations to new kinds of components is good and simple. Adding new operations for existing components is also good, but trickier. Need to examine several existing ASDF systems and extensions to see what kinds of things are being done with the protocol. Plan-then-execute is nasty: makes a lot of things harder, and doesn't buy very much. No need to globally replace ASDF as long as the "future successor" can depend on ASDF systems (and as long as we can extend ASDF to depend on the new kinds of systems.) No project plans were made, but after everyone has had a bit of time to reflect on this, maybe something will happen.