Alan,

AM> I was under the impression that OS reads can/do lock sectors
AM> as they read to ensure consistency in the read. It's a very short time but
AM> it's possible the server is trying to write to this same sector.

Mmm... That would basically mean that the filesystem is a piece of
crap, if it can't handle concurrent access by two processes (a reader
and a writer) to a cluster. It's not like applications (like FBServer)
are let loose to write over disk sectors. :-)

What I can imagine is that, in case a write request cannot be honored,
the file system returns an error to the calling application (Firebird)
which doesn't handle it appropriately and so (in ways obscure to me)
somehow corrupts the database. I can't believe that Firebird
mishandles this situation, so there must be more. Ann, can you confirm
or deny please?

Ciao
--
Nando Dessena
http://www.flamerobin.org