Re: File Notes

El mié, 17-12-2003 a las 15:38, Alexander Larsson escribió:
> On Wed, 2003-12-17 at 19:02, Manuel Amador (Rudd-O) wrote:
> > I can't help to wonder what were they thinking?  If EAs aren't for this,
> > what the hell are they for?  The file system has been claiming our
> > metadata need for AGES, and the kernel people still don't do anything
> > for that.  Evidently, a user-space solution would be easier to code,
> > perhaps, but a kernel-level solution has the guarantee to be more
> > pervasive and easy to use for developers.
> They are designed mainly for ACLs. And due to NFS and other network
> filesystem in common use there needs to be a userspace solution too, so
> duplicating support for EAs might be a waste of work.

If ACLs can be saved, I don't see a reason why small tuples of data
cannot (at least the MIME type, perhaps a comment, anything that can be
regularized and standardized in format and procedure).  NFS already
supports ACLs, and may support EAs in the future.

What the GNOME project needs to do in regard to metadata is to define a
metadata API, call it a lib, and provide at least an EA and a dotfile
implementation.  This would allow for uniform user-level development,
and cover the case for NFS.

Lets put things in perspective:  The fact that NFS doesn't let you share
EAs nowadays is just that.  NFS is a minor case, one that will touch
less than 10% of all GNOME-running computers with EA support.  For that
ten percent, a dotfile implementation of a metadata API, together with
its specification and all related standards, would provide a (perhaps
with lesser functionality) workable alternative, one that would
guarantee metadata existence.

I predict that, in one to two years, 40% to 50% of all the GNOME-running
UNIX and UNIX-like variants such as Linux will have support for EAs,
both at the kernel level and at the user space level (at least the GNU
fileutils, and pax, and tar).  Of those, maybe 10% would use NFS (for
which there would be a dotfile implementation of metadata, just like the
Mac saves its resource forks on a hidden folder in FAT volumes).

This leaves at LEAST 30% of GNOME desktops with EA capability, and no
problem whatsoever falling back to alternate metadata implementations. 
Counting the fact that most Linux distributions currently support EAs
out of the box on several filesystem configurations (ext2, ext3,
reiserfs), you can count on EAs being widely in use in 5 years.

Both of these observations contradict the proposition that an EA
implementation would be useless.  In fact, these observations strongly
support the notion that the EA mechanism/implementation of metadata
should be the default, and the dotfile implementation should be used as
a fallback on volumes where GNOME/Nautilus has detected no EA support.

If there has been a time to standardize and (god forbid this idea)
hijack the EA namespace, it's now, and it should be us.  A standard or proposal should be written around it, and
perhaps a companion proposal should be written for dotfile
implementations of metadata libraries, so both the GNOME and KDE camps
can have their implementation in their chosen languages.

> =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
>  Alexander Larsson                                            Red Hat, Inc 
>                    alexl redhat com    alla lysator liu se 
> He's an old-fashioned moralistic grifter for the 21st century. She's a 
> provocative punk cab driver with an incredible destiny. They fight crime! 
	Manuel Amador (Rudd-O)
	GPG key ID: C1033CAD

Attachment: signature.asc
Description: Esta parte del mensaje =?ISO-8859-1?Q?est=E1?= firmada digitalmente

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