Re: "Modern" server socket programming?

[Marcus Müller]

On 09.01.2013, at 11:17, David Chisnall <address@hidden> wrote:

P.S. I keep pondering adding a kqueue back end for the GSRunLoop stuff, which would be a lot more scalable, but I've never heard of anyone adding enough file descriptors to a run loop for this to matter.

I certainly have no requirement for this either. The pet project I'm currently developing could easily be done in Python as well, but I chose to do it in GNUstep this time in order to see how GNUstep has matured in the meantime.

Interestingly, it has been rather easy this time. :-)

Apart from the problems I already described using ARC in conjunction with GNUstep corebase (which doesn't work) and the few parts in corebase that seem to be missing, the only other thing to note here is a problem with Avahi (via GSAvahiNetService). Using a service name like "D0C0112E-9E1B-4EEE-BD27-B5C3A547D078" (generated via CFUUID) mysteriously raises an assertion in the underlying Avahi library. The workaround is trivial: just use [[NSProcessInfo processInfo] globallyUniqueString], which produces equally well-suited unique ids.

 

Cheers,

  Marcus

-- 

Marcus Müller  .  .  .  http://www.mulle-kybernetik.com/znek/

smime.p7s
Description: S/MIME cryptographic signature