Message-ID: <567@dual.UUCP>
Date: Wed, 6-Jun-84 09:13:37 EDT
Article-I.D.: dual.567
Posted: Wed Jun  6 09:13:37 1984
Date-Received: Thu, 7-Jun-84 07:48:51 EDT
References: <425@hogpc.UUCP> <2727@brl-vgr.ARPA>, <1971@rlgvax.UUCP> <692@cp1.UUCP> <1983@rlgvax.UUCP>
Organization: Dual Systems, Berkeley, CA
Lines: 76

> For years we have described Unix as a very
> powerful software tool kit. I don't know why
> that description should be scrapped. Are the
> users outside the old Bell System that different?

Yes, they are. While many of the *CURRENT* users are probably on a technical
level at least comparable to what exists inside AT&T Bell Labs, the market I 
am selling to, and which many of the other readers of this net are selling to, 
and which represent our mealticket for the next few years, consists of a large
number of people who are not concerned with how easy or elegant it is to
build a solution to a problem, they want the solution ready-to-buy. I don't
like this; I would like to sell to research labs who know as much as/more than
I do, but the economic realities are that this is not the segment of the
potential market that has the money to spend. What is needed to appeal to
these people is not the toolkit, but what the skilled carpenter has built
with the toolkit.

Guy Harris:
> UNIX is a lot of things: it's a portable OS, it's an OS which doesn't make
> life difficult by excessive idiot-proofing (for instance, you're not forced
> to use one of a small set of OS-supplied file formats), and it's an OS which
> comes with a lot of tools and facilities to glue those tools together.
> The third is useful to those shops with the expertise to build systems out
> of those tools *and* willing to put up with the limitations of those tools;
> the second is useful to application developers, and as such indirectly useful
> to the user community; but, frankly, the first is the reason it'll make it
> in the mass market.  If you consider it *only* to be a powerful software
> tool kit, its appeal will be considerably limited.  Remember, few of the
> potential customers for computers are as sophisticated as "we" are.  That
> description shouldn't be scrapped, but it shouldn't be a definition, either.

To my mind, UNIX is a concept, not a product. We have a system that works
tolerably well; the future depends what we can do with the building blocks
we now have. The portability question is important to everybody, but on
different levels. For a hardware company, it means you can build a piece
of hardware, and for a `reasonable' cost you can get a powerful OS running,
without having to invest in a large staff and waiting several years. For the
software developer, it is heaven - a system that allows a piece of software
to be moved to many different vendors' hardware (thus increasing potential 
market greatly) without much trouble.  For the end-user, it means that 
software developed and proven on one machine is likely to run on another.

What is the reality? Well, right now, hardware companies are starting to
realise that they have lost two very important things in going to a so-called
`standard' UNIX system: the lock-in that has traditionally meant continuing
revenues (okay, tell me why a customer would continue to buy my machine if
Bozinski Computers comes with the Roboz-III, which costs 1/2 of what my
machine does, runs the same software, and does it three times as fast. All
of a sudden things like support and expandability matter a whole lot less in
closing an end-user sale), and it also means that the hardware is probably not 
being used at anywhere its' theoretical maximum.

Customers are still fighting all of the hype which says that UNIX is the
be-all, end-all, solution to everything, and finding that the reality is
that even finding a simple solution like a good word processor is much
more difficult than it should be. Have you yet seen a word processor on UNIX
that takes advantage of what the machine can do as well as Wordstar did for
8080-Z80 systems? (Shuts up about NROFF, have you ever tried to support it
on a commercial level for novice users????).

Face it folks, we have a *LONG* way to go. What *I* would like to see are
things like good solid support for VM, so that we could handle things like
shared libraries (bring standard modules in at run-time, instead of at
compile time, (have you ever considered how much disk and memory space is
taken up by the standard `hello.c' program?? Outrageous to waste resources
like that!!!)), and an intelligent way to reconfigure the OS during run-time,
instead of compile-time. Imagine an expandable, object-based UNIX system
kernel, with shared library routines. An operating system should aid in
providing solutions to be commercially successful; so far we only provide
tools.

Thoughts on this diatribe??? Personal flames, are okay....

	    Mats Wichmann
	    Dual Systems Corp.
	    ...{ucbvax,amd70,ihnp4,cbosgd,decwrl,fortune}!dual!mats