[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: ShadowFS (status)

From: Marcus Brinkmann
Subject: Re: ShadowFS (status)
Date: Fri, 3 Aug 2001 15:33:50 +0200
User-agent: Mutt/1.3.18i

On Fri, Aug 03, 2001 at 03:27:33PM +0200, Niels Möller wrote:
> Marcus Brinkmann <> writes:
> > Also an interesting case: One underlying fs provides a file x, and one a
> > directory x.  What do you do?
> read/write requests go to the file, dir-lookups to the directory? At
> least that's a reasonable way to define the "union" of the nodes. I'm
> not sure that it is a reasonable for users, though...
> If you think this way, you never get a "collision" between nodes until
> more than one node supports read/write. One could pervert it even more
> if one node is a read-only file and the other a write-only file...

This is indeed pervers and probably not very helpful.  I also disagree with
the idea that there is no collision.  Obviously, you already have a quite
trivial collision with io_stat().

Anyway, this is a border case.  Some configuration option can make it
flexible to decide if directories should shade files or the other way round,
or if it should be decided by whatever comes first in the list.  Those three
seem to be the simplest solutions to this problem, and anything fancier can
be added later.


`Rhubarb is no Egyptian god.' Debian
Marcus Brinkmann              GNU

reply via email to

[Prev in Thread] Current Thread [Next in Thread]