>> I guess my point that given the architecture dependencies already in AFS, >> we could argue whether or not a "generic" fileserver exists today. I > >pthreads: not ours >rx userlevel: not overly system-specific >namei: generic > >pretty bland stuff
Even given that, there are _tons_ of architecture-dependent #ifdefs sprinkled throughout the code. I still think that calling the fileserver "generic" is a misnomer. >> don't see a huge problem with changing the way we do FD management >> to a more efficient per-platform system. And if it's all hidden inside >> LWP (since that's what seems to handle FDs today; I don't see any calls to >> select() or poll() in the fileserver), does it really matter? > >it's udp. there's a socket. Huh? We were talking about the hypothetical TCP-based fileserver. --Ken _______________________________________________ OpenAFS-devel mailing list [EMAIL PROTECTED] https://lists.openafs.org/mailman/listinfo/openafs-devel
