Coda File System

Re: Theory of writeback

From: Martin Ginkel <mginkel_at_mpi-magdeburg.mpg.de>
Date: Fri, 23 Mar 2007 05:48:08 -0400
u+codalist-p4pg_at_chalmers.se wrote:
> open(,readwrite) triggers store operation at close() as Venus does not
> intercept write() calls and does not know whether the file has changed.

Ok, this might explain something. The files may be opened RW, because
the app sometimes write tags to them. But in most of the cases it
doesn't.

Isn't there some optimization possible? At least on the server there is
some Sha1 calculation on the file's content.
It's difficult to require all apps to be careful with their open().

This might explain also a lot of bogus conflicts, if every
open("x",write) causes the version vector for x to change, even
if no change happens?

	Thx
	Martin
Received on 2007-03-23 18:39:53