[Info-vax] HP wins Oracle Itanium case

[Simon Clubley]

On 2012-08-20, David Froble <davef at tsoft-inc.com> wrote:
> Simon Clubley wrote:
>> On 2012-08-20, David Froble <davef at tsoft-inc.com> wrote:
>>> I really don't understand this attitude that Unix is so great and inevitable. 
>>> Seems mostly a self fulfilling prophecy rather than based upon merit.  I guess 
>>> if you say something enough, people might start believing it.
>>>
>> 
>> Well for starters, in just one market segment, there's getting on for
>> about 1 million new devices per day which contain the Linux kernel
>> (if not the traditional userland tools) been activated by users.
>> 
>> In some markets, Unix is not only inevitable, it has already arrived and
>> it's not going to displaced any time soon.
>> 
>> You will also find, even when the underlying OS is not really Unix, that
>> a number of OS vendors have implemented a Unix-style POSIX programming
>> environment. This appears to be very common in the RTOS world.
>> 
>
> So, are you saying this is what happened, or are you saying that Unix won these 
> jobs by merit?  What I'm asking about is merit.
>
> If DEC had positioned VMS to perform all these jobs, and I'm talking 1990 or 
> perhaps even before, including proper marketing and pricing, would it (VMS) have 
> been feasible, or are you saying VMS could not do the job?
>

VMS as it stands simply could not do the smartphone job I mention above.

The reason is the extreme portability (when compared to VMS) of the
various Unix (and especially Linux) kernels. Porting VMS to a new
architecture is a major multi-year project. Porting the Linux kernel
is still a significant project, but is made _much_ easier by the fact
it's designed to be ported and is mostly (apart from bits of very
low level architecture specific code) written in architecture
independent languages.

When the next new architecture (or major new architecture variant) with a
MMU (and simple user/supervisor mode protection) comes along, it can be
made to reasonably easily run Linux or another portable Unix kernel.
Try doing that with the VMS code base.

The current VMS code base would have to be thrown away and re-written
from the ground up as a portable OS before it could compete in these
market segments to the same extent as Unix. This is part of what the
various microkernel discussions have been about in the last few months
in comp.os.vms when talking about FreeVMS. No sane person would take
the current VMS code base and try porting it to, say, ARM type MCUs.

In addition, the VMS source code would have to be freely available so
that various third parties could port it to their new architectures.
I am not talking about writing a new BSP for a new board/CPU within a
existing supported architecture range here (for which you may or may not
need the source code depending on OS design), but for when you need to
extend a existing architecture or support a new one.

One of the beautiful things about the current Unix/Linux infrastructure
is that anyone can just take the code and port it to a new architecture
or CPU range without having to ask anyone's permission or having to pay
a large amount of money to a vendor to do the job for them.

Also, once you have support for a new architecture, it's easy for a third
party to get Unix/Linux running on various random boards/CPUs implementing
that new architecture without having to go back to a OS vendor to do the
work for you. (This is especially important in market areas such as
smartphones/routers/etc with their product/application specific boards.)
VMS would need this capability as well, even if the source code was
not available.

> I know that DEC blew it by trying to milk every last penny in profits, with no 
> vision of the future.  But what I'm saying is, I don't think Unix was a superior 
> OS, and I do think that VMS could have been the de facto standard that Unix is 
> today.
>
> Address the merits of Unix ..

The merit of Unix is that it's designed to be portable unlike VMS.

In addition, it's API has been specified as a open standard (POSIX) so
a Unix-lookalike interface can be implemented on any other non-Unix OS
you like when Unix itself will not do the job (ie: in a RTOS).

Marketing will not help you if your OS will not easily run on the
next new architecture when it comes along.

Simon.

-- 
Simon Clubley, clubley at remove_me.eisner.decus.org-Earth.UFP
Microsoft: Bringing you 1980s technology to a 21st century world