Some ECLM 2009 Impressions

There's now an official "after-show website" with slides.

I'd first like to thank Arthur & Edi for a very well arranged meeting.

I attended all talks except for one (RacerPro) because I had to get a shot of coffee and some fresh air. This was due to my physical state at that time and isn't meant as a slight against the talk. AFAIK, the ECLM 2009 didn't have an official theme but if there was one recurrent theme it had to be multi-processing / distributed computing in Common Lisp. I'd say in at least four of the seven speeches it was an important of not the main theme of the talk while in another talk it was at least of some relevance.

Here are my impressions from memory since I don't have my notes near me. I will skip the lightning talks for this post. Perhaps someone else will review them.

Dan Weinreb
A Highly-Available Large-Scale Transaction Processing System in Common Lisp

A good talk to start off the ECLM with. Basically a talk about the systems they've built at ITA Software, which CL implementations they use and why, what other languages besides CL are used there and in what systems (and why) and their software building practices (code reviews, don't leave home without them!). Also lots of information on how they're working with the different open source CL implementations they use on fixing bugs and adding features.

Paul Tarvydas
Compiling Diagrams and Graphic Markups for Fun and Profit - a short tour of Visual VDDP and Visual Frameworks

One of the two talks that really stressed the importance of the Electrical Engineering (EE) / Erlang approach to building distributable software components and this talk, like the other talk (David McClain's), also showed a visual interface for making these components and tying them together.

Dimitri Simos
The story of Piano, an aircraft analysis system written in Lisp

A story how Lisp enabled one man to build a software system that's used by a lot (all?) of the big aircraft manufacturers around the world. It was fun to hear how he got his first big customer (Airbus). A very nice talk with a good dose of history, technicalities and advice.

David McClain
GDSP, Butterfly, and Okeanos - a system for visual distributed computing

The second talk about the EE / Erlang approach to building software and a little more explicit about running software on multiple processors or computers. David McClain's website also has some screenshots of their interface. Visually a stunning talk but I'm easily impressed. I have no idea how well it works in practice, but just going by the talk Common Lisp is used at the cutting edge of technology both for its high performance as well as for its features (multi-paradigm, macros, REPL, etc.). One thing that the Free Software hippy in me regrets is that SpectroDynamics's products are not freely downloadable from the internet. It makes perfect business sense ofcourse (to not put it online for free) and I totally respect their decision (I would do the same!) but somehow it feels like a net loss for Common Lisp. Anyway, I need to digest this for a while since I'm probably wrong about it being a net loss for CL. It would definitely be a net loss for SpectroDynamics to just put their products online for free :-)

Kuroda Hisao
A Common Lisp-based Machine Learning Platform

A nice talk about how CL is used in Japan. Something we, or at least I, do not hear much about. Lots of technical information and examples. Also a nice overview of areas for improvement for CL and SBCL in particular. Things which will no doubt be resolved in the near future.

Martin Simmons
Implementing Symmetric Multiprocessing in LispWorks

A straighforward, but good, talk about the work that has been done on adding SMP to LispWorks. Of the commercial implementations LispWorks was the most well represented, not only by several of their employees but also by the amount of people that gave the talks that used it. Of the open source implementations it were SBCL and Clozure CL with SBCL having a clear lead IMHO.


Did Dan Weinreb's talk differ much from his Google tech talk?
Tuesday, 15 September, 2009

I have no idea :) Never seen the latter.
Tuesday, 15 September, 2009

Please watch it and comment. I'll wait.
Tuesday, 15 September, 2009

Xach, I don't have the full 55 minutes to watch all of the talk now nor tomorrow but going by the first 10 minutes I'd expect them to be roughly the same.
Tuesday, 15 September, 2009

"One thing that the Free Software hippy in me regrets is that SpectroDynamics's products are not freely downloadable from the internet."

I believe a previous incarnation of some of the source code was available for educational purposes (to satisfy the Lisp mob's curiosity).

It is no longer available today, but I still have the tarball. Never tried to compile it, though it was interesting to read through. Macros are used to specify each block or filter element like so

(defsigp <white-noise> (<function-block>)
  ((its-name :initform (gen-name <white-noise>))
   (its-descr :initform "WhiteNoise")
   (its-block-color :initform :cyan1)
   (its-noise-state :accessor its-noise-state
                    :initform (make-random-state)))

   (:maker white-noise-source ((ampl 1.0)))

   (:categories :noise :sources)

   (:init-action (gen (ampl 1.0))
     (with-accessors ((state its-noise-state)) gen
       (connect-to gen ampl)
       (insert-function gen
         (fn (ampl)
           (* ampl
              (- (/ (random $MAX-RANDOM state)

Blocks get mapped to classes. You can see the inheritance structure in blocks after blocks.

(defsigp <gaussian-white-noise> (<white-noise>) ...)


(defsigp <fft-block> (<function-block>) ...)

(defsigp <fft-overlapped-processing-block> (<fft-block>) ...)

The distribution doesn't seem to contain a license, most likely an informal release with a "gentlemen's agreement" not to disseminate? So the snippet above is all I'm comfortable sharing I guess. I didn't look much into how the block models are mapped into graphical rectangular blocks on screen, but I have seen many screenshots of those on the web.
Wednesday, 16 September, 2009

Do you know if there are transcripts/papers/audio files from this conference.

I know I sound cheap, but I live far from Europe and some of the talks sound really interesting.
Wednesday, 16 September, 2009

TY: thanks for the information and the link.

kreelman: I don't know if there's anything available but I don't recall seeing anyone with a video camera. I'd just keep a close eye on Planet Lisp and if I were you.
Wednesday, 16 September, 2009

any one record videos of the talks ?
Friday, 18 September, 2009