----- Original Message -----
From: Andrew Phillips <atp@mssl.ucl.ac.uk>
To: <paul@chaos-hovel.demon.co.uk>
Cc: <port-vax@netbsd.org>
Sent: Sunday, August 29, 1999 12:16 PM
Subject: Re: Don't buy a vax, but the vax (was Re: RIP, VAX)


> Hi,
>
> > > To make something that will be successful, it needs to be at a cost
> > > comparable to a PC, and give a little more performance (its a vax, of
> > > course it will!).  Let's not just cater to a select group of people,
who
> > > love the Vax architecture, and don't want to see it die.  Let's make
> > > something that can be used by everybody, and won't cost them an arm
and a
> > > leg, and a third and fourth mortgage.
>
> > I hate to do, this but somebody has to play Devil's Advocate.
> > You may make something that can be used by everybody, BUT, will anybody
> > (outside of VAX enthusiasts) use it? Taking an Analogy in the software
world,
>
> This seems to separate out into two camps, divided along
>    bus lines. PCI vs Qbus.
>
>    Looking at this from a semi-serious perspective, You have to ask
>    what the target market/approach for NuVAX Inc. is. I can see two;
>
>    1) People running VMS that have VAX specific programs not migrated to
>       alpha, that need a performance boost (Sometimes the cost of code
>       migration is enormous, assuming you understand the code enough to
>       migrate it).
>       - This relies on Compaq leaving the market. (Yup, Tick)
>       - You need a running version of VMS. i.e. VMS Source code. (Uh oh)

I thought I read something on the Compaq web site regarding binary emulation
for vax programs running under AXP VMS?

As for PCI support under VMS,doesnt compaq have the code for this in alpha
VMS? It sounds to me though,that what is needed is a bridge between a
supported vax buss and a PCI bus. (the next generation PCI that Intel is
working on BTW is reputed to be 256bits wide,run at the FSB speed of the
CPU,and scale with whatever that speed may be. That would solve the
performance questions between the VAX busses and PCI)

If you wanted me to purchase one of these things,what would you tell me? I
would say to you that I prefer the alpha because its supported by compaq.
Either that or an IBM RS6000. I remember when the R50 blew a disk. We called
IBM and they had someone out in 45 minutes? Can you do that? That kind of
support is exactly why Compaq bought Digital. More importantly even than if
you can do that,how can you convince me that I should put my professional
reputation on the line and recoment that I purchase your product. A similar
situation occured when we needed colour copiers/printers. Many companies
offer good printers. Xerox offered a good system,and teh support that went
along with it. They assured us that they would run in our environment, and
went so far as to say that if it wouldnt work with some of the custom
applicatons that we had,they would go so far as to modify the firmware to
make it work. We knew that once they put this in writing,we could count on
them to make it happen. I felt safe in recomending it. They say noone was
ever fired for buying IBM. Before you get into creating a product,you will
need to be able to answer these sorts of questions.
>
>       Minimum time to market means minimal changes to VMS. This implies
>       that you need to emulate an existing VAX system closely. So no PCI.
>       The easiest target would (IMHO) be one of the "busless vaxes". You
>       want minimal changes so that your mods to VMS are limited to the
>       early boot/SYSLOA images. This means being compatible on the device
>       level with old vaxes. And no economy of scale. You need something
>       quick to exploit this market. I'm dubious as to whether this is a
>       large enough niche to support a company with no track record.
>       The only way to do this is to wait for Compaq to flog off the VAX
>       division, given the complexity of the support chipsets etc.. etc..
>
>       Face it, people aren't going to buy these systems to run *BSD or
>       linux (even if Linux/VAX booted, which it doesn't yet). FreeVMS
>       is even further behind than Linux/VAX.
>
>    2) A VAXpc. I like the microcode idea. Ive been independently toying
>       with the idea of a VAXemulator in a BIOS chipset. Its cheap to play
>       with (there are cheap motherboards with FLASH bios out there) There
>       are VAX emulators in software that could be rommed up to give you a
>       PAL code layer over a PII or StrongARM or Alpha. This is easier, but
>       non-trivial. (Compare the relative thickness of the PALcode
>       description in the Alpha Hardware Ref. Manual for UNIX and VMS). The
>       best chip would prob. be the StrongARM, unless you want floating
>       point performance in which case maybe PowerPC would be better. We
>       dont need 64 bits, and byte moves are expensive on those Alphas that
>       posesss them.
>
>       Advantages of this are commodity peripherals, economy of scale,
>       reasonable performance out of the box, and swift time to market.
>
>       But who would you sell this to? VMS would never run on this beast.
>       or would need severe hacking to do so. VMS talking to an Intel PCI
>       bridge or AGP card? Ugh. There goes the target market.
>
>       We could make the free unixes run on this, but why bother since they
>       probably already run on the underlying native instruction set? Aside
>       from being a "neat hack" of course.
>
>       For this to work, as the world is worrying about 64 bits, it needs
to
>       be price comparable to the alphas/high end PC's and targetted at a
>       specific market.  Unfortunately I can't think of one at the moment.
>
>       I think it unlikely that new application code would be written for
>       this, so would be aiming for the "legacy support" market, which
>       takes you back to VMS and option 1.
>
>    2b) The last option is similar to 2, perhaps using Turbochannel, that
>       just basically admits that this is being done as a hack, like the
>       "freedom CPU" project, throw a design at the wall and see if
anything
>       sticks. If people want to buy NuVAX motherboards that fit into a
>       ATX case that they can plug a selection of cheap commodity PCI cards
>       into then thats cool. Batteries not included. Some self assembly
>       required.
>
>       This is the open source approach rather than a "small company".
>       There _are_ companies selling StrongARM and ALpha based motherboards
>       that do this (Once company in the UK sells/sold a StrongARM board
>       with a choice of Linux or NetBSD as the OS).
>
>
>       Doing a complete system capable of running VMS would have to take
>    the first approach. It also risks the fate of the BeBox, which although
>    a great design, didn't quite sell enough units. To make a new VAX
>    based system you need a compelling advantage over the default.
>    "The only actively developed VMS platform" is one, although perhaps a
bit
>    short term. Others: A 64bit VAX processor? One that smokes an Alpha?
>    (or my favourite - one with 8kb pages?). Get real.
>
>       IMHO The only way to do this is option 2b. Open source targetted
>    at some mass produced hardware, with the option of plug in PCI boards
>    to do Q-BUS or what ever. "Download this file and Flash into your bios
>    to get a VAXpc! Click here for ASUS, here for TYAN, here for SOYO..."
>
>       The worst thing is that, if, by some cosmic act of supreme
>    wierdness, this became a going profit making concern, Compaq would
>    probably want a slice, unless you'd worked out a firesale IP deal
>    beforehand.
>
>    I can't believe I'm seriously discussing this. I must be nuts.
>    Time for another beer.
>
> andy.
>
> --
> atp@nojunk-mssl.ucl.ac.uk             |        Dr. Andy Phillips
> phillips@nojnk-isass1.solar.isas.ac.jp| Mullard Space Science Laboratory
> a.phillips@nojunk-ucl.ac.uk           | "It's the late 1990s, This is a
spam
> atp@nojunk-coralcay.demon.co.uk       | protected .sig. You know what to
do"
>