Re: Continuations for the JVM proposal

[Jim White <jim at pagesmiths dot com>]

Per Bothner wrote:

> ... 
> Certainly, continuations are appealing for web-pages, and are
> useful at the source code and design levels.  But they don't
> need to be implemented by the JVM, though of course that improves
> performance.

Performance (along with simplicity and ease-of-use) is the whole issue 
here AFAICS.

One of the great tragedies in the Java saga is that the ad hoc threading 
system was thrown in (which consequently had to be deprecated) rather 
than a continuation-based API which would have had meaningful formal 
semantics along with the rest of the JVM and which could have simplified 
the JSR-133 issues.

Of even more practical consequence is that the Java threads 
implementation in most JVMs (like Sun and Kaffe) actually use 
heap-allocated stacks that could be trivially exposed as continuations 
for use by Scheme and other sensible languages.

So while Kawa having a good continuation compiler for currect JVMs is 
necessary, I think it would be also be a good thing if you supported 
this prospective JSR and made sure it did the right thing so Kawa would 
benefit.

Jim