Re: [Setup-tool-hackers] RE: 2.4: System Tools - Solaris

[Sander Vesik <Sander Vesik sun com>]

On 30 May 2003, Luis Villa wrote:
> 
> I don't think you're asking the right question, Murray. Portability has
> to be like all of our other guidelines- it's not 'is it accessible' it's
> 'can it be made accessible if someone puts the work in.' 
> 
> So, if Sun has legitimate architectural critiques, then that's a
> showstopper. If Sun just says 'we don't care and won't port it', that's
> not. We can't limit ourselves to the set of features someone is willing
> to port to Solaris, because, frankly, there aren't many Solaris-based
> volunteers.
> 
> Sander[1] does not appear optimistic about the architecture[2], but I'd
> love for someone else to take a look. Joe Marcus's preliminary work is
> encouraging to me on this score.
> 
> Luis
> 
> [1]Who is never optimistic about anything ;)
> [2]I don't think the design critiques are valid- it would be very nice
> to have an utterly pluggable design as he's suggesting, but we have to
> be realistic about the resources available. If someone were to write
> what he's suggesting for gst2, of course, that would be great. :) 
> 

I don't think the matter is utter plugability - the support of gnome on
its / tops20 / zOS isn't exactly wide - but having the basic ability to
discover what the relevant dataset is on the OS at hand and behaving
appropriately. This does not require some hypothetical total plugability
of everything, but merely pretty simple level of plugability - essentially
not straightjacketing the backends (for which you need the ui to be
slightly dynamic).

The example with user information containing 'login class' is a very
simple exmaple and comes from freebsd (which started the thread), and the
feature has been around and active for many years by now (it comes from
4.4bsd). I don't think it matters whetever you can manage the login
classes themselves from gnome-system tools, but the gnome-systems tools
should lose information about the users class and should expose that
information. Presently this has no (reasonable) support. 

And this is not one isolated arbitrary case - there are similar problems
for network protocols other than IPv4, network devices that do not emulate
ethernet MACs, network device speed and media being a reported and
settable parameter, etc - the problem is extensibility beyond the systems
and configurations that the authors have, not support for some not
unixlike at all system.

	Sander

OpenOffice.org - conquering the world 14000 PC-s at a time