Log of the #nice channel on irc.freenode.net

Using timezone: Central European Time
* ArtemGr joins03:22
* arjanb joins11:55
* bonniot joins14:41
<arjanb>hi14:49
<ArtemGr>salute
I'm trying to write my first OCaml program... What do you think of F# + IKVM combination, BTW?14:57
i will probably need a fast "system" language to implement webserver around the Java app, thus i'm evaluating OCaml. possibility to integrate it with Java application on a Mono platform considered as a big plus..15:05
<arjanb>that's an exotic combination15:14
is F# anything more than a proof of concept?15:15
<ArtemGr>F# is pretty much working, from what i've read, though Mono isn't directly supported. There is also OCamIL wich is going to have Mono support and full OCaml compatibility (it is a port of OCamp compiler with it's optimizations).15:18
<arjanb>interesting though i prefer haskell as functional language15:22
<ArtemGr>last time i've checked haskell performance, it was inadequate, unfortunately.15:24
and it seems to be 'too purely' functional for a practical language...15:27
<arjanb>yeah haskell has a different programmer speed/ program speed trade off15:28
<bonniot>you think Java/Nice would be problematic performance-wise?15:37
<ArtemGr>Much of the load on web application is handling of slow TCP/IP connections. The best approach here is thread-less - using the "poll" function on a list of opened sockets. While NIO might help there, i can't be sure it will perform good. Java is poor in handling strings, too much copying from place to place and form method to method. Looking at NIO implementations doesn't help in denounce this assu15:50
mption.. Having a separate thin web-server achieves other goals as well, such as balancing requests between different implementations (stable and experimental), recording "real workload" for later testing and hiding JVM crashes from the end user. So, since separate process is required anyway for these things, i thinking about implementing it in a top-fast, more stable then Java, and interesting la
nguage intead of dull C/C++... Java stability for the moment is very poor. Sun Java VM fails on FreeBSD and (seldom) even on supported platforms, and other platforms were problematic even to compile, recently...
s/and other platforms/and free VM implementations/15:51
<bonniot>who is developing the FreeBSD vm?15:52
<ArtemGr>a group of volunteers. they patch Sun initial releases on their own.15:55
for example, NIO is most probably is implemented over "select" (i've checked some time ago), and "select" is a) slower (requires slow array manadgement of sockets) b) amount of sockets is limited; "poll" must be preferred, on FreeBSD at least (becouse on some platforms poll is silly implemented over select)...15:58
* bonniot leaves17:55
* ArtemGr leaves20:07
* arjanb leaves20:23
* ArtemGr joins20:37
* ArtemGr leaves23:21