Richard Sims wrote:
> [other wrote:]
> >It could actually be much worse than that, as we have discovered.
> >The default behavior of Archive is to FOLLOW SYMBOLIC LINKS.
>
> This is as it should be.  The intent of archival operations is to preserve
> images of data at given points in time, to serve versioning, legal, and
> other needs.  A symbolic link is only an allusion to data and just doesn't
> fit the paradigm.  The purpose of Archive is to capture data at any given
> instant in time, and that's what it does.  It does not capture symbolic
> links or directories (though it remembers the directory structure for
> basic reconstruction upon Recall).
>

begin SOAPBOX.STANZA

Richard, I think your interpretation of 'the praradigm' is narrow.  The
'image' of the data, it seems to me, should preserve the data as it
exists.
If that means that it preserves a symbolic link then so be it.  After
all
a symbolic link is simply a special file whose contents specifies the
name
of another.

In fact, the very fact that a symbolic link can point to a file that
does not
exist (a fact that has nothing to do with the existance or functioning
of ADSM)
exposes the falicy of ADSM's treatment of symbolic links.  It behaves as
if
you've deleted a file under its feet, even though nothing in the
filesystem
changes from the time you requested the archive.

You talk about 'image'.  If what is retrieved does not match what is
archived,
then it is not an image.

Philosophy and semantics aside, the practical reality is that ADSM's
treatment
of symbolic links in archives is in the least inflexible (why not allow
the
option to preserve the link as a link?) and at the most dangerous.
Witness
the very fact that the Web SMIT fileset on AIX 4.3.1 contains a circular
link
(as is allowed by the filesystem) and the fact that ADSM archive decends
ad
infinitum.

Try opening a problem on that:
ADSM support: 'ADSM is working as designed, why would those boneheads
put in
a circular link?'
AIX support: 'AIX is working as designed using an allowed configuration,
you
should talk to those boneheads who don't treat symbolic links
correctly?'
I paraphrase; I'm sure they didn't use the word bonehead.  I'm sorry,
though,
but I agree with AIX support.  Look at 'tar'.  That's tape-archive.
That's
_archive_, in the same sense as was meant by the designers of ADSM (it
is only
with 'find . -mtime -3' that tar behaves like ADSM's backup).  Yet tar
treats
archives the way it should.

Practically speaking, I would rather, as a user of ADSM, be held
responsible for
ensuring that my file specification includes 'left field' when I've put
a link
that points to left field than having to write a ridiculously
complicated script
that runs multiple archives which avoid getting symbolic links so that
the data
isn't copied multiple times.

And speaking of 'the way it should work' (in my humble opinion), why
can't we have
the option of crossing filesystem boundaries?  I respect that when I
archive
'/* -subdir=yes' I only get stuff in the root filesystem, but wouldn't
it be nice
be able to say '/* -subdir=yes -crossfsboundary=yes' and have it archive
the
entire system?  Or '/application/* -subdir=yes -crossfsboundary=yes' and
have it
archive all the data in the filesystems /application/home,
/application/db,
/application/exports, etc.

Oh, and why not allow 'exclude.archive /tmp/.../*' to be valid syntax
for the
include/exclude list?  The allow '-archignoreexcludes' in the same way
that we
now allow '-archmc' to override the include/exclude list for includes.

The backup processing includes code to figure out what filesystems are
mounted
and backs up 'all-local' except /tmp.  You never miss when a new fs is
created.
Why should I have to duplicate said code in a script to achieve the same
result?
Isn't code reuse an honourable goal?

And...
I could go on.

I've been working with ADSM intensely for nearly three years and learned
from a
master who knew it in beta version 1 and influenced its development.  I
love the
depth and breadth of what ADSM is and does.  However, some things are
just done
plain _wrong_, or, in the least, inflexible, with the _option_ missing
to do things
the way I (and a lot of others) feel is right.

end SOAPBOX.STANZA

Cheers...
Bruce