Mikola Lysenko wrote:
> Part of the problem with CSP-style languages is that implementing a non-deterministic choice operator or 'guarded' select is that it is very difficult.  Frankly, the version in DCSP is rather slow and unnecessarily limited.  In the future, I hope to address these issues, but I won't be able to do any work until I finish up my current projects.

The Newsqueak papers talk about the implementation of “select”; the Plan 9 thread(2) or Russ Cox’s libtask have the source code for the “Alt” structure and its related code.  It’s a solved problem.

--Joel

renoX wrote:
> That said I thought that the content of the presentation is a bit 'old': after all he was presenting 20+ old ideas.

Twenty-year-old ideas that /still/ haven’t gotten their fair use except in one obscure (but wicked cool!) OS and one fairly weird and still obscure embedded processor—anyone do any Transputer work recently?

Read Pike’s paper on the implementations of 8½ & rio; these are powerful ideas that programmers deserve to have available to them.

--Joel

Limbo has a ref keyword that makes a ref to an adt so you just copy refs around not huge copies of the data structure.

I've done a postgresql client in Limbo, it's great having the CSP co-routines just chat away to each other. It really separates out the way you approach a problem.