Darren New wrote:
Andrew Lentvorski wrote:
Huh? That's not true. NFS has locking.
I've never seen a version of NFS that even guarantees that what you
write eventually gets to the server, let alone locking. Maybe they've
fixed that in the last few years.
NFS works just fine under Solaris and has since something like 2.5.
That's what, 1996 or so?
NFS locking also has worked just fine since FreeBSD 5. I happen to have
been the person who rewrote the rpc.lockd for FreeBSD 5. This was back
in 2000-2001.
In addition, while I like to slag on Linux for crappy NFS
implementations, I believe that even *Linux* got this right. Quoting:
http://playground.sun.com/pub/nfsv4/webpage/nfsv4-wg-archive-dec-96-jan-03/1665.html
For what it's worth, modern Linux systems use a kernel-level NFS server
which correctly implements the syncing semantics demanded by the NFS V2
specification. (NFS V3 support is currently underway but its not yet
completely supported.)
NFS has close() before open() semantics. It also obeys fsync() semantics.
Just because crappy programmers do not understand the semantics of
fsync() or close() is not the fault of NFS.
-a
--
[email protected]
http://www.kernel-panic.org/cgi-bin/mailman/listinfo/kplug-lpsg
