OS costs

[Dick Dunn]

In response to my whining about ongoing development staff...
> > ...  Instead, the development is nearly continuous, which means
> > current sales have to support a full-time development staff...

...jes at mbio.med.upenn.edu (Joe Smith) writes:

> While I don't doubt for a minute that this is currently true, doesn't
> this strike anyone else as completely outrageous?...

Yeah...as a matter of fact, it does!  It contributes to the end-user cost
of the product in a big way.  I would argue that the market is forcing this
level of development, in two ways:  demand for support for new hardware,
and "feature madness."  The need to support new hardware will always be
there, and will always require a certain ongoing development cost.  We
might be starting to level out a bit...it's not like in the earlier days of
PCish UNIX where only a few devices were supported.  At least we've got a
reasonable set now.  The additions are more in the class of "would like"
and "should have" than "must have."

What about the feature madness?  It looks to me like the system producers
(the folks responsible for generating the various source bases) are con-
stantly trying to one-up each other.  How do we get out of that mode?  It's
an important question, because it bears on Joe's next point:

> ...  If Unix is
> ever going to 'mature' as a product, it's got to shed the need for
> intensive system hacking, and the needless duplication of effort on
> the part of vendors ...

As the systems keep adding features and keep growing, it gets harder and
harder to maintain them, let alone add anything.  Everything interacts. 
That's a formula for less-stable (i.e., "immature") software.  Trouble is,
people see UNIX (but not DOS or OS/2) as the platform for future develop-
ment.  A lot of ideas get tried out on it, and some of them stay.  Not all
of the ones that stay belong there, but that's the way it goes.  Beyond
that, just enough of the new features are of general interest/need that
people clamor for them and they get rushed into "production" releases.
I'd sure like to see a way out of that, somehow decoupling experiments from
commercial systems.
-- 
Dick Dunn     rcd at ico.isc.com -or- ico!rcd       Boulder, CO   (303)449-2870
   ...I'm not cynical - just experienced.