PDA

View Full Version : Musing about standards (small follow-up) ...



H.j. Woltring, Fax/tel +31.40.413 744
05-25-1992, 08:34 PM
Dear Biomch-L readers,

The following note was received today from the Family Medicine and HSPNET-L
lists; I am X-posting it because of the pertinence -- generally speaking --
of its contents to our own 3-D Kine(ma)tics Standardization debate, even
though the underlying issue itself is rather remote from our interests.

Regards -- hjw.

- = - = - = - = - = - = - = - = - = - = - = - = - = - = - = - = - = - = -

From: fam-med@gac.edu 26-MAY-1992 09:18:39.34
Sender: Hospital Computer Network Discussion Group and Data Base

Comments: Resent-From: Donald Parsons MD
Comments: Originally-From: stpeters@dawn.crd.ge.com (Dick St.Peters)

Probably of interest to all interested in whether to make their network
UNIX with X-windows or MSDOS -- Don

----------------------------Original message----------------------------

A couple of weeks ago Dr. Faughnan asked, and I finally have time to reply:

> >The technical world has passed that hurdle. A workstation comes with
> >networking built in - heterogeneous multi-vendor networking. But we
> >are still struggling mightily about standards. Not that we don't have
> >them - we have lots of them, several for most any imaginable interface :-)

> Which of these nascent standards do you think would be worth looking into?
> Can you recommend any clear writing on the topic?

These are tough ones. It's not hard to find clear writing, but it is
very hard to find unbiased writing. Getting the technical community
to agree on standards is often like trying to get Catholics, Jews, and
Muslims to agree on a common theology. Anybody who has looked closely
enough at a standard to write clearly about it has most likely done so
because he or she has a stake in the outcome.

The good news is that the outcomes of most of these battles matter
little to most end users. However, they can be virtually life or
death struggles for vendors, especially small ones dependent on the
success of some innovation or too small to absorb the costs of
retooling when a standards decision goes against them.

With the caveat that I have my own biases - so consider yourselves
warned - let me write a bit about a standard that I think does matter
to the end user. It's POSIX, from the IEEE Portable Operating System
for Computer Environments Committee.

The operating system interface is the single most important interface:
when moving an application - or a user - from one computer to another,
the iron rarely makes any difference. It's moving applications or
people from one OS to another that creates all the headaches.

One solution dating to about 1970 is to have a portable operating
system, one that can run any machine. This was part of the original
rationale for UNIX, and today you can get UNIX to run everything from
Crays and mainframes to low-end PCs to digital telephone switchgear.

Because POSIX builds on UNIX's two decades' experience in implementing
a portable interface, POSIX is often called standardized UNIX. Now,
I'm a UNIX advocate, and the thought of being able to say that UNIX
won the OS wars is very appealing. In the technical community, the
big OS war has been between VMS and UNIX, and being argumentative by
nature, I've never hesitated to wade in for a good fight. That war is
especially fun for me, since I'm one of the few people with extensive
experience with both operating systems.

But on a list like this, I have to behave more responsibly and admit
that POSIX doesn't mean UNIX won. In fact, it doesn't even matter
whether the OS is portable or not. (You won't tell my technical
friends I said that, will you?) Only the interfaces (human and
programming) to the OS need be portable.

The current POSIX standard is a first step in standardizing the
interface between applications and the OS - the programming interface.
A number of operating systems that are very definitely not UNIX comply
with the current POSIX standard, or will soon. Most vendors with
non-UNIX operating systems have signed up to have their OS's track
future POSIX standards.

Part of the reason is that POSIX is a minimal standard. It does not
require that an existing interface be removed or changed, only that a
POSIX compliant one be added. This is a key point. While I don't
overflow with sympathy for vendors who, in my view, failed to see the
UNIX light after two decades of opportunity, leaving their customers
high and dry would be quite another matter. A lot of my fellow UNIX
bigots fail to understand this.

The longer term intent of POSIX is to move the industry toward
standardizing as much as possible of the OS interface, at a pace slow
enough to minimize disruption. Because almost all the research that
has been done on portable OS interfaces has been within the UNIX
framework, POSIX is certain to continue gradual adoption of a
standardized UNIX programming interface.

Recently, it looked like a new OS war was shaping up, pitting Windows
NT vs. POSIX-compliant OS's (mostly UNIX implementations). However,
battle, in that context at least, was temporarily avoided when
Microsoft agreed to make NT conform to the current POSIX, because the
world's largest buyer of computers, the US government, essentially
absorbed POSIX into the latest FIPS (Federal Information Processing
Standard).

Microsoft is non-committal about whether they will comply with coming
POSIX standards though. My opinion - remember, I'm biased - is that
they face a dilemma. If they endorse POSIX, then they become just
another player, another entry in the race to see who's implementation
of POSIX is best. Effectively, NT becomes just another UNIX. If they
do not endorse NT, then they're betting on being able to set a de
facto standard, a replay of the DOS scene, that can stand up to POSIX.

My biased opinion is that they can't do another DOS. Too much of the
PC market is large companies for whom portable software is becoming
virtually a Holy Grail. I think a lot of software vendors see it this
way too, and that's why we're seeing so many PC applications appearing
for UNIX - the vendors don't want to bet their future on a company
with its own non-standard programming interface and its own non-
standard window system (user interface).

Which brings up the other significant standard, The X Window System,
to give it its proper name, or just X. I won't say nearly so much
about it, because I don't like it. However, it's a standard that's
there, so I use it - and spend a lot of my time having to program
around it's limitations.

One thing I will say about X is that it gets too much credit from
inexperienced people for things that are inherent in any halfway
decent windowing system. Most of the things people ooh and aah about
are things workstation windowing systems have been doing for a decade.

--
Dick St.Peters, GE Corporate R&D, Schenectady, NY
stpeters@crd.ge.com uunet!crd.ge.com!stpeters