I found most of the talks interesting and some of the speakers were directly outstanding. I am not in violent disagreement with the other reports, but I will try to list some items I found noteworthy. In this, I will most undeservingly skip a lot of interesting and fun stuff, for which I apologise in advance.

Sunday was dedicated to tutorials. Like many others, I would have wished I could have gone to all of them. I ended up with the webapp tutorial Edi Weitz gave on Hunchentoot and Duane Rettigs tutorial on optimization in Allegro CL. I found Hunchentoot interesting and my impression is that it is a pretty well engineered piece of software which is little surprising given Edis trackrecord. The Allegro tutorial ended up being more ACL specific than I had hoped but it did offer a good insight into the mechanics of ACL and Duane is a good presenter. Some good stuff is on its way in ACL 8.1, that's for sure.

The conference started for real Monday. One thing that struck me was the creativity with respect to presentation layouts. As I normally move around in the corporate world and Microsoft PowerPoint presentations, it was and inspiring to see how much could be done, also with freeware tools. PowerPoint was represented, but mostly in the Mac instantiation. I saw very few people running windows. Some presenters used quite some time moving back and forth between their talk and their demos. More people should probably have done like Chris Connolly who did a talk on the FREEDIUS imaging system in using a virtual desktop to streamline the changing back and forth. Chris used Virtue on his Mac. I have since downloaded it to my own Mac and it works really well.

Robert Strandh gave a talk on ESA (a part of McCLIM) and the future of Emacs like applications in Common Lisp and CLIM. While I like the prospects of having a stronger programming language behind the editor, I am also somewhat worried that his idea that one should pack things into separate applications will move ESA and Climacs in a direction away from Emacs that will make porting from Emacs to Climacs (and similar tools) difficult. He did demo a mail client and while it was neat that such a thing exists, it probably has a long way to go before it is as powerful as GNUS. He did say at one point that he uses Climacs and the mail client every day. We will see how ESA will affect the world.

Antonio Leitao spoke on translating Java libraries into Common Lisp. It was based on the analysis that Java and Perl has many more libraries than CL and that many of the existing CL libraries suffered from the 80% problem in that they would develop to the point where the cover the authors needs rather than going all the way. The first part is obviously correct, the latter I am less sure of, certainly numerous exceptions exists, but with a much smaller number, semidead projects are of course much more noticeable. His approach was interesting but obviously also one that demanded significant manual intervention. A thing one might wonder was how well the difference in number of libraries reflect actually deficiencies in the Lisp world; perhaps the effect lies less in the numbers as such and more in the size of the particular markets leading to some high-quality options.

Christian Queinnec talked about Computer Science education. He is a very good speaker, funny and interesting, and they are clearly doing a lot of good stuff at the Université Pierre et Marie Curie. They have automated a big part of the grading in the undergraduate education which among other things means that they can go back in time and study the growing corpus of examination and exercise answers and for instance retry the grading part. In the questions part, he gave a simple recipe for comparing texts for similarity. Gzip each text separately and the gzip the concatenation of them. If the latter is much smaller in size than the sum of the individuals, a big overlap must exist!

Michael Sperber gave an enticing report from the Scheme R6RS process. He confirmed that standardisation is a very hard and painful process (he claimed it to be the most painful thing he has done since the breakup with his first girlfriend), a quite apropos comment to the ongoing debate on comp.lang.lisp. He also said that the R6RS process has abandoned the old consensus principle the other revisions employed in favour of being able to move forward at all. There will also be a community process once the committee has finalised the report which incidently will consist of two parts, one for the language and one for a standard library. My favourite item in the whats new part has to be the addition of Serious IO :-)

Tuesday started with a talk on CL-HTTP. I really liked the talk, unlike others who found it weird to various degrees. One thing that struck me was that CL-HTTP seems less dead than I thought it was. There are some issues with the license, and I understand that the attitude of "nobody has been sued yet" is not good enough for a commercial user, but I cannot really see CL-HTTP as being in a whole different situation than so many other open source projects. The authors concentrate on some things and leave other things behind. Apparently the license it not so easy to change but remains good enough for John Mallory so I think it to be a bit unfair to blame him for not prioritising the issue. I have definitely decided to check it out.

From the talk on SC, a set of SEXP languages that compiles to C, I noted the comment that C is useful as an intermediate language but it lacks access to the execution stack, something that the SC system solved as I understand it.

The talk on social network analysis (CL-SNA) clearly demonstrated that there is a lot of theory on the subject. It is not quite my field, but I did wonder if some of the analysis apparatus would not be useful to telecom operators trying to understand the traffic in their networks. The speakers also revealed that the next European Lisp Users Group meeting would be in Istanbul. The idea is good, I would very much like to see Istanbul and Turkey being the tourist it is means that travelling there should not be completely out of reach. We shall see; if anybody said (or has posted elsewhere) when it is, I have missed it.

Cyrus Harmon has elsewhere been applauded for a good talk. I read in one of the other reports that it was done using LaTeX which surprises me, given the high quality of the layout. Most of the software he has developed for his PhD is available; I am going to check out his gcc-xml-ffi in my long and dusty search for a good FFI tool.

From the talk/demonstration of LispWorks, I picked up a very neat idea: step-to-cursor. Perhaps nothing new, but it was to me. LispWorks incidently offered all conference attendees a 30 days evaluation copy of LispWorks. I picked one up and look forward to trying it out.

Richard Jones gave a good talk on Dynamic Memory Management and the challenges in garbage collection. I noted the mentioning of Azul Systems that makes custom hardware that somehow supports garbage collection in a way I have now forgotten. One of the thing Richard said was that the Lisp community, having invented and forefronted GC, should make sure hard earned lessons from the Lisp machines and other places are not forgotten and left to reinvention in modern systems design.

Wednesday started with a rather impressive talk on and demonstration of HOP, a web programming framework built on top of a homegrown Scheme implementation. Manuel Serranos presentation was all coded directly in HOP and it seemed really neat. I spotted Erick Gallesios name on the frontpage. I really liked his STk scheme implementation that used many years ago; I believe these guys know what they are doing even if the 200-300 KLOC code base sounds a bit scary.

I found the talk on the OWL reasoner interesting. A few good points I noted down: description logics are about possible worlds - OO is about one particular world and RDF triples is the cons cell of the semantic web. Incidently, their product (Racer Pro) uses CL-HTTP. Mentioned was also Wilbur.

The talk on XMLisp was interesting. The idea is to modify the reader to be able to use real XML with angle brackets and all directly inside the lisp code. To me it seemed to improve readability but at the cost of typing where the conventional SEXP based form should be easier to deal with (and better supported by editors such as emacs). XMLisp would translate XML trees into objects which some sitting close by found a very unnatural representation but perhaps not if one is into the DOM/SAX way of looking at things. Certainly it showed of just how powerful the Common Lisp reader is! Alexander Repenning also seems to be doing very interesting work with respect to teaching children about programming, a most important issue.

I had looked forward to Jans Aasmans talk on scalable Lisp and was no disappointed. It might have been better of as a half tutorial rather than a 50 minutes talk but a lot of good stuff on dealing with very big data sets was presented. It appears that Franz has been deeply immersed in the ACM library searching for techniques for their AllegroCache and AllegroGraph products. Their was some talks during the question session on the value of cons-free functions which went somewhat over my head.

Tags: ilc07, lisp