> I could quibble - they have to do all that anyway, right, across the
> various
> clients?
>
> I don't think so. Consider, how would you map an ACL for a UNIX file in
say, an AIX DFS filesystem, to an ACL for a Windows file system? The client
is supposed to honor those ACL's, and restore them along with the data. I
don't see how that would work.
And you can have things in non-text files that simply don't compute when
moved from UNIX to Windows and vice-versa. Moving a binary from UNIX to
Windows won't make it executeable. And there are file formats that have
big-endian vs. little-endian issues (for example, output from Fortran, for
which there were UNIX compilers at one time, although I realize there are
fewer and fewer such cases these days.).
And Richard is correct, when restoring to Windows, TSM doesn't write the
files; it calls NTFS to write the files. I don't know that all the
filesystem calls translate. For instance, you can create an empty 2GB file
in an NTFS filesystem; there is no such thing in a JFS.
And there are at least 5 different filesystem types for AIX, at least 4 that
I know of for Red Hat, et cetera, et cetera.... many, many permutations.
SO if you say well those aren't the files I'm interested in, I just want
flat files...then I think what you are looking for is a different protocol
altogether - maybe ask for a "bit stream restore", where TSM will restore
the bits into a flat file across operating systems, ignoring all the
conversion issues, and make you responsible for decoding the result.
But I'm still not sure that's desirable from a security point of view, and
most people will find it less work to restore the file back to its original
O/S and FTP it, thereby taking the responsibility of stripping out the ACL's
themselves...
W
>
> Lindsay Morris
> Principal
> TSMworks
> Tel. 1-859-539-9900
> lindsay AT tsmworks DOT com
>
>
> On Tue, Jan 5, 2010 at 2:22 PM, Richard Sims <rbs AT bu DOT edu> wrote:
>
> > On Jan 5, 2010, at 12:19 PM, Lindsay Morris wrote:
> >
> > > True, unix to unix works, and windows to windows too. Osx and netware
> are
> > oddballs, but not so common.
> > >
> > > But we build an appliance that wants to "dsmc restore" from ALL the
> nodes
> > on a TSM site, windows and unix both, to spot-check recoverability. Dsmc
> > blocks cross-platfrm restores (windows vs unix at leadt), and for no good
> > reason, AFAIK.
> >
> > Well...consider what's involved. The good reason is that a fully
> > cross-platform client architecture would have to include full programming
> > for the universe of file system types accommodated on all the supported
> > platforms - and additionally would have to adeptly provide a meta layer
> of
> > emulation of the wealth of system calls associated with the file system
> > which is not native to the receiving platform. That's an enormous amount
> of
> > development and testing and maintenance. I'd rather that their energies
> > went toward things like 64-bit clients, and administrative API, and
> > long-overdue evolution of the terribly neglected CLI.
> >
> > Richard Sims
> >
>
|
