problem with oplocks.

[Olaf Frączyk]

On Tue, 2003-02-18 at 03:34, Michael B. Allen wrote:
> On Mon, 17 Feb 2003 15:02:59 -0600
> > Um... Just curious, but how are "oplocks are unreliable by definition"?
> 
> I wondered what was meant by this too. I concluded it was just a zealous
> choice of words. I believe he means that a) even after being granted an
> oplock break the client may still find the file is locked and ultimately
> get a sharing violation and b) on any system other than Windows or systems
> with kernel oplocks the file can still be written to and possibly c)
> if the oplock holder looses connectivity and another writer commits
> changes data will be lost. There's nothing unreliable or technically
> flawed about the protocol though. NFSv4 will have the same issues.
Hi,

Sorry if I had confused you.
I was thinking about the "c" scenario.

Regards,

Olaf Fraczyk