NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/os/os2/advocacy/10946

Internet Message Format | 1992-12-26 | 7.2 KB

Path: sparky!uunet!paladin.american.edu!gatech!concert!ais.com!bruce
From: [e-mail] (Bruce C. Wright)
Newsgroups: comp.os.os2.advocacy
Subject: Re: OS/2 bigot meets NT....
Message-ID: <[e-mail]>
Date: 26 Dec 92 01:29:34 GMT
References: <[e-mail]>
Distribution: usa
Organization: Applied Information Systems, Chapel Hill, NC
Lines: 123

In article <[e-mail]>, [e-mail] (Timothy I Miller) writes:
>
> In article <[e-mail]>,
> [e-mail] (Robert Suh) writes:
>>
>>Well, I can name lots of reasons...
>>
>>1) C2 Security - I know lots of places that Security of data on the PC or
>>LAN is a MUST.  Can't have someone just walk up to a PC/Node and take a look
>>at all the data on it.
>
>      I'm pretty trusting of people I let use my computer, or hook up
> to it for that matter.  The only computer I'm going to be hooked up
> to will be the other computer in my house, and I don't really need
> to keep the files secure from my girlfriend...

Most of the places I know that actually _need_ C2 level security (as
opposed to just some kind of security to keep users out of each
other's hair) are either Government installations or Government
contractors, and some of them of course need higher levels of
security than that.  Unfortunately the way the economy is going they
may turn out to be the only big computer users left :-(.

>>2) Cross Platform - Wouldn't it be nice to be able to sit in front of either
>>a PC, MIPS, Alpha, or almost any other system (Thanks HAL ) and know that
>>Windows NT has the ability to run on it?  I know thats a BIG Plus for some of
>>the larger mixed - environment workplaces.
>
>      This whole cross-platform thing is a little confusing.  From
> what I understand, only the OS is cross platform, and the programs
> must all be recompiled in order to run on them.  So, let's say I
> have a lot of money invested in NT software for Intel machines.
> Then I decide in the future that I want an Alpha machine.  I'll have
> to go out and buy all of my software again.  Shouldn't I have an OS
> that's designed to take advantage of the special properties of the
> hardware instead?  It would end up costing the same, and my OS would
> end up being more powerful.
>      Of course, if the same software could be used, things would be
> much easier and cheaper, but too many abstracting layers would need
> to be used in order for the system to be useable.

My understanding is that NT on platforms like the Alpha will be able
to run Intel (DOS/Windows) programs in an emulation mode, but that in
order to take advantage of the faster machine you will need to recompile
the application.  This, of course, is old wine in a new bottle -- Unix
and VMS have had various 3rd-party products that do a pretty good job
of this now, but they're *slow*, which you'd expect from an emulator.

There are two approaches that they might take with this:

    1)    Make an emulator for NT like the one for Unix/VMS.  This has
    the advantage of being relatively straightforward, since the
    emulator could probably be largely adapted from existing code.
    It has the disadvantage that apps would be very slow with it;
    and why would you buy an Alpha if you're just going to throw
    away all the performance advantages it has by running old Intel
    code in an emulator?  It seems to me that this will only be
    interesting for apps used `occasionally' that you can't or don't
    want to relicense for the Alpha;  if you really need the all the
    performance of an Alpha for one or more of your critical apps
    you're going to require that those apps be in native mode.

    2)    Use a machine translator like the VEST translator that can be
    used to move VMS code from the VAX to the Alpha.  This can do much
    better than an emulator, but is still not as good as directly
    compiled code for the target machine (the penalty is probably
    only a factor of 2 or 3 or thereabouts rather than a factor of
    10-20 or more for emulation).  This is problematic;  even apart
    from any licensing issues of whether your license agreements for
    the translated software allow you to run it on the Alpha, the
    VEST technology pretty much requires that you understand some-
    thing about the internals of the program being translated, in
    order to select the proper translation mode.  The Intel .EXE
    file format and procedure calling standards (or lack thereof)
    also gives the translator less information for its job, so it
    may not be possible to do as well with Intel images as it is for
    VAX images ... if it's possible at all.

A machine like the Alpha is not targeted at the typical PC user, or even
the typical PC `power user' -- it's targeted at users that require very
large amounts of compute power, but don't necessarily require very much
PC compatibility.  I can see why it might be interesting for such users
to be able to run a few non-critical, non-compute-bound PC programs that
give them access to programs that they like, but if they run everything
that way it defeats the purpose of the faster CPU.  Their main app(s)
really _must_ be native apps, otherwise why pay for the Porsche?

Which gets back to native mode applications for NT.  I've heard various
rumor-level talk to the effect that not many NT developers plan to target
anything but the Intel platform.  (I know it's not _all_ of them, I know
of a few that will be targeting others as well, including Alpha).  If it
turns out that there are few native-mode packages available for NT on the
Alpha, I suspect that most Alpha users will use either Unix or VMS --
which is probably the OS that was on their previous workstation anyway(!).
This is the sort of thing we won't really know until NT is out -- it's
not even something that is very well within Microsoft's control:  about
the only app that they control that might use that much CPU power is Excel,
when running unusually large spreadsheets;  W4W is not the type of app
that you need that amount of CPU to run.

But in any event NT won't be getting a black eye because of its size, as
Steve Withers suggested in another article -- Unix and VMS are both in
the same general ballpark as NT in that regard.

It's an interesting question.  NT's future on Intel boxes looks to be
considerably brighter in the short run than its future on non-Intel boxes,
for what that's worth (I'll pass for now on the question of how bright
NT's future on Intel boxes is).  It will depend on whether it offers the
end-user anything not offered by Unix or VMS, in terms of basic function-
ality and app support.  Most of the `advantages' listed over those systems
are advantages for the _developer_, not the _user_;  which means that, if
we're talking about purely technical issues, it will probably be a long
time, if ever, before it has much effect (example:  MS-DOS has a horrible
API from a developer's perspective, but it's been very successful in spite
of it).  And NT has some substantial _deletions_ from the feature set
enjoyed by Unix or VMS -- full multiuser support (eg, on serial comm
lines) for one thing.  IMHO, the biggest thing NT on non-Intel boxes has
going for it is Microsoft's marketing strength, but it's unclear how well
that will translate into the high-end workstation world.

1993 will definitely be an interesting year.

Bruce C. Wright