[Shootout-list] Weekly Update
Isaac Gouy
igouy2@yahoo.com
Fri, 24 Sep 2004 10:35:17 -0700 (PDT)
> >> They are aimed at showing off some of Erlang's strengths by
> >> requiring lots of message passing (in the case of the new
"ringmsg"
> >> test) and loading of plugins (in the case of the new "plugin"
> >> test).
> > imo adding a test to show-off the strengths of a language is not
> > keeping the spirit of honest comparison. It's simple advocacy.
> Isaac, you are misinterpreting what I said. I am not stacking the
> deck in favor of Erlang. Instead, I know that Erlang was designed to
> solve a particular category of problem... If they cannot, that
exposes
> a weakness in their implementation, just as string processing...
Sorry, it wasn't real hard to misinterpret. (iirc Joe Armstrong
suggested the ring at LL)
We quietly dropped the requirement that producer-consumer use kernel
threads - doesn't that test cover context-switching?
The extra thing the new test examines is process/thread creation time.
Creation-time is general enough to apply to a reasonable number of
languages. Once we insist on massive concurrency, the test becomes very
narrow; instead fork a thousand processes and tear-them-down n-times.
That mirrors the object tests: one test for messaging, one test for
creation.
_______________________________
Do you Yahoo!?
Express yourself with Y! Messenger! Free. Download now.
http://messenger.yahoo.com