Re: [Gnu-arch-users] Re: How does arch/tla handle encodings?

[Jan Hudec]

On Sat, Aug 28, 2004 at 14:15:50 +0000, Esben Mose Hansen wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On Saturday 28 August 2004 10:11, Jan Hudec wrote:
> 
> > > 1. Filesystem support for extended attributes (many file systems support
> > > this) 2. Standard utilities support for extended attributes (cp copies EA
> > > etc., tar stores EA, etc.)
> > > 3. A standard for specifying encoding in EA.
> > > 4. Make diff, patch and friends aware of this and act accordingly.
> > > 5. Make tla store these extended attributes and changes together with the
> > > files and changessets.
> >
> > About every sane editor writes the file by:
> > 1) Writing to a temporary file
> > 2) Renaming it in place of the old one
> > (yes, it is sane -- it's a unix "poor man's atomic commit")
> > ... and where are your precious extended attributes??
> 
> Either copied over from the original file, or lost, dependent on the editor. 
> Did I mention 2) above? I do not think, however, that this is tla (or arch)'s 
> problem. Thus, I propose to ignore this bit.

You are right. It's not arch's problem and it will actualy get settled
quite fast once EA interface settles down.

> > Things that are thigs tightly coupled with file's contents should better
> > be part of the contents. You want cat pass them and cat, by desingn,
> > does not, and have no way to, pass extended attributes around. Extended
> > attributes may be used, but as a last resort where nothing else is
> > possible, not as a standard.
> 
> :) Some will not agree with you. Some will. I could write a long text about 
> why, where and how you are wrong, but I find this irrelevant in this context.

Yes. There are reasons for both ways. And in fact, there are probably
situations where one or other is better. So I really think neither
should be forced upon applications and users.

> More interestingly, do you disagree about 5) above? Should tla handle EA 
> "sanely", or simply throw them away? THAT is the interesting discussion in 
> this forum. I, for one, believe EAs are here to stay. The support for them 
> are slowly trickling up&down the toolchains. I would think i prudent to at 
> least consider how tla and arch should handle this.

I actualy agree with 5. EA's are reality, so tla should support them.
I however think that:
1) The interface has to settle down a bit first (it's likely the xattr
   interface will be obsoleted by file-directory hybrids)
2) GNU tar has to support them first (may be it does, but I don't know
   about it).

-------------------------------------------------------------------------------
                                                 Jan 'Bulb' Hudec 
<address@hidden>

signature.asc
Description: Digital signature