On Tue, 25 Mar 1997, der Mouse wrote:

> > Actually, Irix's package tools are quite well documented...
> 
> Then the documentation is hidden well enough that neither I nor whoever
> "[had] to agree with [me]" above found them, which amounts to much the
> same thing.

The Online Developer Insight Docs contain a 200 page volume on the
packaging tools...

> > You can find the manual for them online at my old employer, BTW...
> > check out:
> > 	http://santafe.arch.columbia.edu:88/SGI_Developer/
>
> I don't think it's reasonable to point to columbia.edu web pages as
> support for a claim that SGI's tools are well documented.  How is the
> random SGI sysadmin supposed to know to look on columbia.edu (on a
> nonstandard web port, at that), even assuming Web access?

This is just the standard online documentation exported through a web
gateway _which_is_bundled_with_every_copy_of_Irix_.

> 
> In any case, I looked there.  Which link are you referring to?  The
> only one that looks at all promising is the "Software Packager User's
> Guide", and it talks about packaging software for swmgr, whatever that
> is, not inst.  Even ignoring that, it is user-level documentation for

swmgr is the graphical counterpart of inst.  At any rate, a quick poke
at the search field at the top of said page turns up substantially more
information.

All of this is beyond the point, though...  My intention was never that
we _duplicate_ the format and exact appearance of Irix's packaging
tools.  Before being leapt upon for bringing up a strong point in
an SVR4 based UN*X, my point was, as it is now that the package tools in
Irix are a good example of an interface which meets many of our current
needs, including but not limited to:

	* Support for multiple architectures in each package

	* clean separation of components of each package, with
          well-designed dependencies, and clear distinction between
          MD and MI parts

	* nice graphical, term-menu, and command line packaging and
          installation tools.

	* ability to install anywhere in the tree, and to install by
          copying or sym-linking.

I maintain that we could learn a lot from looking at this system...

> the SGI tools; it is not, as far as I can see, documentation on any of
> the issues I cited as being undocumented (the major ones, IIRC, being
> the format of software package files and the inst-private (or
> swmgr-private, as the case may be) area where it keeps records).

If you are interested in more information on these issues, feel free to
reach me off list...  Again, the exact implementation details of Irix's
tools are not relevant to the issue at hand.

> Perhaps the IRIX package tools _would_ be a good choice...but I still
> stand by my claim that they are complicated and undocumented, and
> maintain that those qualities (though the latter more than the former)
> disqualify them from serious consideration here.

To repeat:  Irix's package tools are a good example of an interface we
would do well to implement in a manner compatible with our own system.
They already address many of the issues we are seeking to deal with in a
clean and elegant way.

--
				Jim Wise
				jim.wise@turner.com