motives for AT&T, etc (was re: att & osf)

[Donn Terry]

There has been a series of postings on this topic all with the theme that
"somebody has to make UNIX(TM) 'right'".

The "somebody" isn't a single *vendor*, nor is it the standards bodies,
but rather the marketplace.

The existence of standards (such as POSIX) makes it reasonably possible
for buyers to specify what they want in a vendor independent fashion.
(Just like nuts and bolts.)  They then can choose based on price,
additional features, whim, or whatever else they like between the systems
meeting that standard.

More importantly, if the standard isn't what they want, they can tighten
up the specification.  They can also add to it, or even change it.  In
doing so they take a risk of having no vendor meet the specification, or
a higher price, or delays, but it's the buyer's risk, and if its
valuable enough to him, he'll take it.

A purchaser like the Federal Government (in the form of a FIPS) has done
exactly that with the POSIX FIPS.  It's tighter than IEEE's POSIX standard,
and the size of that market will draw most vendors to comply with the 
FIPS (and thus with IEEE's version).

You want compliance to a standard:  you'll have it within probably a
year from most vendors, and it was done by free-market means, not via
contractual enforcement or other coercion.  (A vendor doesn't have
to comply with the FIPS; it just might want to however; the income is
nice.)

A proprietary specification could, theoretically, have had the same
effect.  The several attempts I know of have not had that effect.  One
of the reasons is that because they weren't created by an open process
and weren't (perceived of(!!)) as reasonably independent of the owner's
whim, and thus competitors weren't that interested in follwing the
specification, possibly to their own disadvantage.  More importantly,
those specifications do not appear to me to be technically attractive
enough to overcome this problem.  Had they been, they might have
succeeded in spite of the above problem.

In the case of OSF: there is a baseline specification that they will deliver
(read the literature).  It's not really well defined yet, but that should
happen.  If that specification matches buyer's needs, what OSF has will
sell.  If it doesn't, OSF goes out of business.  In this case, the direct
buyers are the vendors of systems.  If OSF's specificaitons meet the 
vendors' customers' needs, the vendors will buy it and pass it thru, adding
what's missing for their particular customer base.  If not, the vendors
will solve the problem some other way.

The folks at OSF aren't stupid, and they realize this.  To the extent that
existing specifications match the marketplace, they'll adopt them.

The best way, in my opinion, to assure the success of the proprietary
standards (to the demise of today's leading open standard, the UNIX
System) is to try to legislate the "right" answer, rather than letting
the customers decide in a free market what they want.

Please don't delude yourself into thinking that somehow, magically, you
know what the customers want.  You, at best, understand *what you heard*
from those customers you talked to.  That may be a factor, but I've seen
many a product be of more value to an unintended marketplace than for
the target market for which it was designed.  (I think UNIX's target
market was originally amateur astronomers, at least according to one
story I heard.)

Donn Terry
HP Ft. Collins

My comments represent only my own opinions, that of my employer or anyone else.