On Mon, Jul 16, 2001 at 11:44:54AM -0500, Dimitri Maziuk wrote: | * D-Man ([EMAIL PROTECTED]) spake thusly: | ... | > Interesting ... from my understanding UDP is a connectionless | > protocol, and as such packets aren't guaranteed to arrive at the | > destination. It seems to me that for a file system, you really want | > all the packets to arrive. How is this not a problem? (BTW, I'm just | > beginning in networking programming and have no NFS experience yet, | > but don't be afraid to give gory programming details :-)) | | IIRC the idea was that you don't want to use NFS over slow and unreliable | internet links anyway, you'll only use it on the LAN. Packet loss isn't | much of an issue there; OTGH, performance penalty caused by TCP overhead | is, esp. if you're trying to e.g. swap to an NFS drive. | | It's a reasonable trade-off, given that there were no gigahertz CPUs or | gigabit ethernets. With modern hardware it's the other way around: we'd | rather have reliable transport and a [usually negligible] TCP overhead.
Ok, that makes sense. How about if probability leaves us behind and a packet is lost? Does NFS provide any way to correct for that or will your filesystem be hosed? -D
