[OpenAFS-devel] Re: Breaking callbacks on unlink

[Andrew Deason]

On Tue, 28 Feb 2012 12:46:34 -0500
Jeffrey Altman <jaltman@your-file-system.com> wrote:

> On 2/28/2012 12:22 PM, Andrew Deason wrote:
> > And how does this make the callback break a waste? If the contents are
> > gone, forgoing a callback break can (and does, as we've seen) result in
> > inconsistent state between clients.
> 
> Inconsistent state between clients was not the issue that was raised
> here.  The issue was unpredictability of continued file access.  The
> designers made a compromise decision.  Do not break the callback on
> the last unlink because it will expire on its own and by doing so
> permit other clients that are using the file a bit more time to
> continue using what is in the cache.  That is all.

Yes yes, I'm not arguing about the decisions made in the past. What it
sounded like you were saying is that any such callback break is
pointless and serves no purpose (a waste), with which I disagree. It's
not what we do and there are reasons for that, but it's not like, pure
overhead.

> >> Go back and re-read this thread.  Polling was brought up in
> >> discussion yesterday in the exchanges between Troy and Simon.
> > 
> > The last post I see from Simon is from Saturday. Post a link or a
> > message-id or something?
> 
> Simon's message from Saturday describes polling but does not use the
> word.

As far as I can tell, it was just mentioned as something that could
hypothetically be done (and undesirably, at that). Your earlier message
sounded like a response to the suggestion that something does that
today.

-- 
Andrew Deason
adeason@sinenomine.net