[OpenAFS-devel] Re: Breaking callbacks on unlink

[Andrew Deason]

On Tue, 24 Jan 2012 17:40:18 -0500
Jeffrey Altman <jaltman@your-file-system.com> wrote:

> Runtime configuration is not an option in this case because the
> behavior of whether or not a callback is broken in response to an RPC
> is part of semantics of the RPC.
> 
> The way you change the semantics of an RPC is by implementing a new
> RPC.  It is by implementing a new RPC that the client and server are
> able to agree upon the semantics.

As a general statement, I think that heavily depends on the behavior in
question (to what degree it is just a bug). For this situation, sure, I
can see that.

It's not immediately clear to me whether this is a semantic implied by
the RemoveFile RPC or the ones that register the callback in question
(FetchData, etc)... for this immediate issue the point is moot (since
I'm not going to try to alter the protocol to accomodate). But for the
future solutions to this, that seems relevant.

-- 
Andrew Deason
adeason@sinenomine.net