[Info-vax] Reimplementing VMS, was: Re: HP adds OpenVMS Mature Product Support beyond the end of Standard Support

[JF Mezei]

On 14-02-04 01:45, Johnny Billquist wrote:

> And I disagree. While 64-bit was pretty much a requirement, the MP 
> developments are (in my opinion) pretty much irrelevant here. These are 
> implementation details for different chips, and new stuff can happen 
> there without much change to the codebase. What happens to already exist 
> in silicon is pretty irrelevant to the decision to port or not.

The installed base for BCS systems need "mainframe" class systems with
efficient memory system which the 8086 lacked. AMD did have a compelling
product and in 2004 Intel countered by announcing not only a move to 64
bits, but also to the CSI memory system which would put the 8086 on
equal footing with IA64 for memory, and better footing for CPU.

So in 2004, the message was fairly clear. And while HP's contractual
commitment to IA64 may have lasted a few more years, HP could have
started a porting project back in 2004 to be ready by the time the 8086
grew up and gained Quickpath/64 bits.

But my gut tells me that LaCarly's ego did not want to admit IA64 was a
mistake and continued to expect IA64 would be great eventually, until
Intel knocked at the door and asked to put IA64 out of its misery.

By that point, if HP had not looked at porting to 8086, they would have
been forced to do so.

> But moving back to a 32b-bit architecture would have been very unlikely, 
> so until x86-64 started to spread, there was not much to look at.

One needs to consider lead times for this. A pilot to evaluate
feasability may be quick, but full porting takes much longer. Look at
how quickly the VMS guys managed first boot of VMS on one of them IA64
things. The rest of the work took quite a bit more time.

If the commitment to IA64 ran out in 2007, (with last chip released
then), it would mean that you would have to be ready with your ported
OSs by then so that new customers get the 8086 version right away, while
existing customers take their time to move, during which they can buy
the last back of IA64 servers.


If HP didn't plan it this way because they refused to see IA64 as a
failure, then by end of original contract, HP wouldn't have been ready
to move to 8086 and woudl have been forced to pay Intel for longer IA64
life.

One possibility was that the first contract was meant to extend IA64
with Madison, Tukwila, Poulson, Kittson giving HP enough time to port
and start selling 8086 based BCS systems before end of life on IA64.
Perhaps at some point in time, HP changed its mind about porting and
instead decided to abandon BCS alltogether. This would explain the new
contract in 2010 to spread the remaining generations of IA64 and
dismantlement of VMS engineering.

one would have to go back to see when Livermore started to make the
"won't be ported beyond Itanium" rather blunt statements. This would
indicate at what point the door was closed to porting to 8086 with the
plan to let BCS die of old age.


Normally, a person of that stature within corporation would have made
neutral statements such as "we are constantly evaluating what happens in
the market, but we are very happy with the performance of IA64". When
livermore started making her "won't be ported beyond IA64" statements,
it was a airly clear message that HP had decided against the ports.


> port VMS. But given how much work it would be, it's not something that 
> ever made sense from a business point of view until rather recently. And 
> rather recently also means a much smaller custom base, which cannot 
> support such a port anymore.

Nop. VMS' fate was definitly sealed irrevocably sometime in 2009 when
Hurd signed the order to get rid of VMS engineering and hire a handful
of indians to do the maintenance work. No serious porting would have
been considered beyond that because the product was already written off
has having no future (and no engineering to do the work).