[OpenAFS-devel] Re: Breaking callbacks on unlink

[Andrew Deason]

On Mon, 27 Feb 2012 11:10:06 -0600
"Matt W. Benjamin" <matt@linuxbox.com> wrote:

> But just to be clear, what exactly would be the proposed semantics at
> the different clients?  Ie, using existing RPCs and callback
> registration, or using new RPCs you would propose (to afs3-standards
> ;)?

I can't speak for Troy, but at least what was in my head is that the
wire is all the same; just that when the fileserver gets a RemoveFile
RPC, if there are callback promises active on the file (or just
'always'), the vnode doesn't actually go away until X minutes after the
last callback goes away.

I suppose there's also potential issues of that there's no way for a
client to really verify something is gone (for either space purposes, or
confidentiality or something). I'm not really trying to propose actually
doing this or anything, though; it just seems like from talking with
chas that it's the best that can be done without altering wire protocol.
Whether or not that's acceptable I'm not really dealing with; just
sharing some ideas :)

-- 
Andrew Deason
adeason@sinenomine.net