using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Alex Zarochentsev]

Hello,

Does anybody think that using linux extended attributes API would be good for
accessing specific reiser4 file attributes?

Currently all that stuff is exported through pseudo files under "metas"
subdirectory.  It might be good for scripting but it leads to security problems
in legacy unix applications and it requires having +x permission on regular
files or modifying linux VFS.  

It is not a problem to have two API for accessing file attributes, XATTR 
and pseudo files.  At least XATTR will be usefull in places where security
matters.

Regards,
Alex.

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Chris Mason]

On Mon, 2004-04-05 at 07:54, Alex Zarochentsev wrote:
> Hello,
> 
> Does anybody think that using linux extended attributes API would be good for
> accessing specific reiser4 file attributes?
> 
> Currently all that stuff is exported through pseudo files under "metas"
> subdirectory.  It might be good for scripting but it leads to security problems
> in legacy unix applications and it requires having +x permission on regular
> files or modifying linux VFS.  
> 
> It is not a problem to have two API for accessing file attributes, XATTR 
> and pseudo files.  At least XATTR will be usefull in places where security
> matters.

We've discussed this in private before, but I'll add my thoughts to the
public discussion.  

xattrs are the current linux api, you'll make it much easier for people
to use reiser4 by supporting it (in addition to whatever new and cool
API you develop).

-chris

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Hans Reiser]

We are not going to use xattrs, get that through your head.

Hans

Alex Zarochentsev wrote:

>Hello,
>
>Does anybody think that using linux extended attributes API would be good for
>accessing specific reiser4 file attributes?
>  
>

>Currently all that stuff is exported through pseudo files under "metas"
>subdirectory.  It might be good for scripting but it leads to security problems
>in legacy unix applications and it requires having +x permission on regular
>files or modifying linux VFS.  
>
>It is not a problem to have two API for accessing file attributes, XATTR 
>and pseudo files.  At least XATTR will be usefull in places where security
>matters.
>
>Regards,
>Alex.
>
>
>  
>


-- 
Hans

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Narcoleptic Electron]

Hans Reiser wrote: 

> We are not going to use xattrs, get that through
> your head.

Perhaps it would be more constructive to point him to
an FAQ on the subject instead of being rude.

If there is no FAQ entry on the subject, and you don't
want to bother writing one, then expect these messages
regularly.


______________________________________________________________________ 
Post your free ad now! http://personals.yahoo.ca

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Alex Zarochentsev]

On Mon, Apr 05, 2004 at 01:54:14PM -0400, Narcoleptic Electron wrote:
> Hans Reiser wrote: 
> 
> > We are not going to use xattrs, get that through
> > your head.
> 
> Perhaps it would be more constructive to point him to
> an FAQ on the subject instead of being rude.

:)

I am a reiser4 developer, there is no FAQ for the developers yet :)

I just wanted to give a constructive alternative to the "metas" discussion in
this list.

Thanks.
-- 
Alex.

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Alex Zarochentsev]

On Mon, Apr 05, 2004 at 08:55:56AM -0700, Hans Reiser wrote:
> We are not going to use xattrs, get that through your head.

If people found xattr support useful in reiser4, someone would implement it,
not necessary one from reiser4 team.

> Hans
> 
> Alex Zarochentsev wrote:
> 
> >Hello,
> >
> >Does anybody think that using linux extended attributes API would be good 
> >for
> >accessing specific reiser4 file attributes?
> > 
> >
> 
> >Currently all that stuff is exported through pseudo files under "metas"
> >subdirectory.  It might be good for scripting but it leads to security 
> >problems
> >in legacy unix applications and it requires having +x permission on regular
> >files or modifying linux VFS.  
> >
> >It is not a problem to have two API for accessing file attributes, XATTR 
> >and pseudo files.  At least XATTR will be usefull in places where security
> >matters.
> >
> >Regards,
> >Alex.
> >
> >
> > 
> >
> 
> 
> -- 
> Hans
> 
> 

-- 
Alex.

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Alexander G. M. Smith]

Chris Mason wrote on Mon, 05 Apr 2004 08:46:26 -0400:
> xattrs are the current linux api, you'll make it much easier for people
> to use reiser4 by supporting it (in addition to whatever new and cool
> API you develop).

One of the big points is to cut down on the number of APIs and gain
in saved effort.  Listing attributes?  Just use good old ReadDir().
With a common name space, you can use grep and other well developed
tools to deal with attributes.  And so on.

- Alex

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Hubert Chan]

>>>>> "Alexander" == Alexander G M Smith <agmsmith@rogers.com> writes:

Alexander> One of the big points is to cut down on the number of APIs
Alexander> and gain in saved effort.  Listing attributes?  Just use good
Alexander> old ReadDir().  With a common name space, you can use grep
Alexander> and other well developed tools to deal with attributes.  And
Alexander> so on.

But if programs are using xattrs for other filesystems, then people
don't want to have to detect when they're running on Reiser4, and switch
to using ReadDir() et al.  So while the philosophical side of me says
that using normal file operations is the better thing to do, the
pragmatic side of me says that there should be at least some sort of
compatibility layer that uses the xattrs API and makes the appropriate
translation for Reiser4.  (And whether or not that really is the best
thing to do runs along the same lines as whether or not WINE helps
migration from Windows to GNU/Linux.)

-- 
Hubert Chan <hubert@uhoreg.ca> - http://www.uhoreg.ca/
PGP/GnuPG key: 1024D/124B61FA
Fingerprint: 96C5 012F 5F74 A5F7 1FF7  5291 AF29 C719 124B 61FA
Key available at wwwkeys.pgp.net.   Encrypted e-mail preferred.

Re: using of XATTRs instead of pseudo files for accessing reiser4 file attributes.

[Chris Mason]

On Mon, 2004-04-05 at 18:41, Alexander G. M. Smith wrote:
> Chris Mason wrote on Mon, 05 Apr 2004 08:46:26 -0400:
> > xattrs are the current linux api, you'll make it much easier for people
> > to use reiser4 by supporting it (in addition to whatever new and cool
> > API you develop).
> 
> One of the big points is to cut down on the number of APIs and gain
> in saved effort.  Listing attributes?  Just use good old ReadDir().
> With a common name space, you can use grep and other well developed
> tools to deal with attributes.  And so on.

Sure, I think namesys has some great ideas on the new pseudo files.  

But, there are existing apps using the xattrs, it makes sense to get
them for a relatively small programming effort in reiser4 instead of
having to modify each one and get patches up stream.

-chris

Re: using of XATTRs instead of pseudo files for accessing

[mjt]

Chris Mason wrote:

>But, there are existing apps using the xattrs, it makes sense to get
>them for a relatively small programming effort in reiser4 instead of
>having to modify each one and get patches up stream.

As Alex Zarochentsev pointed out, he just wanted to present an alternative
in xattrs and that someone would implement them if they were considered
good.

Hans Reiser (a bit rudely, if you don't mind me saying, but I'll
say it anyway) shot this down.

I haven't had time to investigate this matter, but is it possible to
map the Reiser4 API onto the xattrs API? I may be way out of my league
here, but isn't the Reiser4 API mostly a better ("better") xattrs
implementation?

A transparent layer in between would probably please most parties,
and if Namesys isn't interested, someone else would do it, no?

Reiser4, as superb as it is, is pretty borderline to some standards.
I think it paramount to acknowledge this and minimize upstream patches.

Upstream patches may piss off a lot of people, especially in cases
where they have written something minor for themselves and compensating
for this is more work than what seems ok.

IMO, Reiser4 should be able to work just as any other fs in those cases
when it's needed, without having to worry about the sexy details
which people will use in other cases.

-- 
mjt