xfs
[Top] [All Lists]

Re: to compare journalised file systems

To: Ivan Rayner <ivanr@xxxxxxx>
Subject: Re: to compare journalised file systems
From: Matteo Centonza <matteo@xxxxxx>
Date: Thu, 17 Oct 2002 09:53:42 +0200 (CEST)
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: <20021017105957.6d3af7f2.ivanr@xxxxxxx>
Sender: linux-xfs-bounce@xxxxxxxxxxx
Hi,

On Thu, 17 Oct 2002, Ivan Rayner wrote:

> On Wed, 16 Oct 2002 14:15:24 -0800, Ethan Benson wrote:
> 
> > On Thu, Oct 17, 2002 at 12:04:00AM +0200, Andi Kleen wrote:
> > > > xfs already honors an extended attribute for files only, the
> > > > maintainers believe that honoring such a thing for directories is
> > > > evil.
> > > 
> > > If 'the maintainers' really said such a thing then I disagree with
> > > the maintainers. Honoring it for directories should be fine and I
> > > can think of applications for it.
> > 
> > fine look up the list archives if you don't believe me.  when the
> > change was first commited Ivan used the word evil specifically when
> > referring to nodump on dirs.  there was a more detailed discussion a
> > few monthes later, thier concern appears primarily security.
> 
> By the "maintainers" you mean me, since I was the one that implemented
> it.
> 
> I don't remember using the word 'evil' -- obviously I was just
> grandstanding.
> 
> It wasn't done for directories for 2 main reasons, that I recall:
> 
> 1. Security.  I didn't like the idea of the owner of a directory
>    deciding whether files, that might be hidden in the depths of
>    a directory tree and could be owned by others, get dumped or
>    not.  To allow this would be ... evil! :)

as i argued before

http://marc.theaimsgroup.com/?l=linux-xfs&m=102370432004601&w=2

security concerns are blown away if you thing this as administrator stuff 
(via a reserved namespace using attributes). 

> 2. Simplicity and performance.  To expand this for directories in
>    xfsdump would mean a fair bit more work (which would not have
>    been approved as a project), and probably a real performance
>    hit during dump processing.

These are both real concerns. Regarding performance, probably you can 
trigger the extra directory pruning stage or whatever else via a flag to 
xfsdump.

Just speculating ;).

Anyway, thanks the very good work.

-m


<Prev in Thread] Current Thread [Next in Thread>