Monkey Face@unixprt (email@example.com)
12 Aug 87 04:22:13 GMT
> The claim of statelessness should not be taken too literally. When one
> of our disks got trashed (@*#% Xylogics!), we did the "obvious" thing -
> we took the server down, created the filesystem afresh, restored all of
> the backup tapes, and then let the clients continue where they left off
> before the crash. Talk about some confused clients...
> "Stateless" only means that the state is kept on disk rather than RAM.
Also, NFS is implemented in a stateless manner, but the other network
services that are provided are very state-full. Since NFS is really just
a 'base' for file system services, extensions via servers et al, are and
will continue to be state-full (i.e file locking). Several other network
services will be state-full also.
The above example is an interesting case when this approach falls a little
short, although in general, for short periods of 'failure' it is very
This archive was generated by hypermail 2.0b3 on Thu Mar 09 2000 - 14:38:49 GMT