Mark writes exactly what I would have written, and thus I'm going to repeat
it all here again verbatim. Nonetheless, to help make the situation perhaps
just a little clearer, let me also repeat a diagram I included in my first
posting:

                  deep
              philosophical
                firewall
                    .
      MPE apps      .        Linux apps
         |          .            |
         |          .            |
      MPE O/S    <- . ->     Linux O/S
         |          .\           |
         |          . \          |
         |_____________\_________|
         |              \
     Linux kernel        \ file sharing would be
                           accomplished through
                           file equations/links

Let me also point you to the following slides that are on Intel's website
concerning the ARIES binary emulation of PA-RISC/HP-UX on the IPF (Itanium
Processor Family)/IA-64 family of processors. Of the forty slides, I believe
these two are the most important:

     http://www.intel.com/design/itanium/tranhp/sld009.htm
     http://www.intel.com/design/itanium/tranhp/sld024.htm

although the slides begin at:

     http://www.intel.com/design/itanium/tranhp/sld001.htm

The right-hand column of slide 9 represents the totality of the reasons for
generating a binary compatible PA-RISC emulator. The performance levels
achieved vs. native PA-RISC performance appear in slide 24, which is
summarized elsewhere as approximately 30 to 70% of PA-RISC's native mode
performance.

If you have the time, I'd recommend looking at all of the slides, of course.
After looking at the slides, the primary question that comes to mind of
course is why wasn't this option offered CSY? Booting HP-UX can't be all that
much different that booting MPE.

Wirt Atmar


======================================

Matt Shade ponders:
> <To preface this, I just want to say that I'm trying to find
> answers for my own peace of mind. I hate sounding like an old
> stick in the mud, but I've been having problems coming to
> grips with reason behind the cause. I'm also sure I'll get
> flamed big time for this, but I need to get this off my chest.>

I cannot imagine anyone flaming you for such a well-reasoned and thoughtful
post.

> Perhaps I'm just dense, or maybe behind a little in the
> reasoning, but I just can't see the point in an OpenMPE
> project based on an emulator. Unless it was purely for a
> teaching tool, or a hobby, or a toy. I don't see an emulator
> running on top of another OS (linux) running a company's business.

Well, if I understand Wirt correctly, it is not running on top of Linux, it
uses the Linux kernel.  This is just the most basic of operating system
services.  The kernel does not include the file system or a shell for
example.  The point is to use parts of Linux that gives us the portability
and saves us the time and trouble of writing things like device drivers.  By
doing this, we avoid the pain that CSY is experiencing.  I do not imagine
booting up Linux and then typing MPE.  When you boot, the system would mount
some MPE volume sets and start the PA-RISC emulator as a system service.  As
with any Lunix machine, your shell would then start.  Of course our shell
would be the CI or MPE shell.  This shell would do all of the same functions
that the CI would, system variables, file equations, etc.  The exception
comes when one goes to run a program.  If the program is an MPE binary, the
PA-RISC emulator executes the program.  If it is a regular Linux program,
Linux runs it.

> I keep picturing some MIS director who's been sweating for a
> couple years, wondering how to migrate all his financial and
> manufacturing apps off his HP3000s, saying "Whew! Thank god!
> Now I can just buy a couple high-end PCs, load up linux, and
> then install MPE on top of them, and I won't have to migrate
> to a new OS and I can keep my job!"

Good, that's what I envision too!  (Well, maybe with less sweating...)

> Yes, that's probably a ludicrous example, but I'm not sure
> why there would be an OpenMPE if not for those cases.
> Otherwise, what's the point?
>
> <snip> But I
> can't see how, for example, a company could honestly justify
> moving from an HP3000 with 300 users to a PC (or Sun box, or
> HP9000, or whatever) running an MPE emulator on top of linux.

I can't see how a company could honestly justify moving from an HP3000 with
300 users to Windows NT.  ;-)  But keep in mind that Linux runs on much
bigger boxes than just high-end PCs 9as well as NT fo the record).  Maybe
even Superdome... ;-)

> HP has had a hard enough time expanding or even keeping it's
> HP3000 customer base. So what makes anyone think that
> creating an MPE emulator will change that? Or is that even
> the reason behind OpenMPE? Is the reason just because WE love
> MPE so much and don't want to see it just fade away? Are we
> hoping some underground movement will catch on, and soon
> every linux distribution will include an MPE emulator? (I
> KNOW that's not the case, but again, I'm just trying to get a
> feel for it)

Right now we are in an either/or situation.  Either you run your apps on the
HP3000 or on Linux.  I think the Santa Fe proposal gets you to the point
where that isn't an issue anymore.  Maybe I have a app that runs great on
MPE for one job but not another.  Now you can run an MPE app and Linux apps
on the same machine at the same time.  Now we've killed another rap against
the 3000, "I don't want to maintain multiple hardware platforms".  As for
attracting new developers, you don't have to convert someone to do only MPE
work.  WHEN Image is ready, it will be a welcome change for those who put
their data in dbm (indexed) files.  IBM is already trying to get users to
use their journalized file system.  Is the MPE file system better than
IBM's?  So, yes, I do see a possibility of getting more users on the MPE
bandwagon.

>With regards to the feasibility of an emulator, most of you have seen what
I wrote in
>reply to Wirt's proposed roadmap for OpenMPE.
<snip />
> But, IMHO, the amount of performance
> that would be sacrificed would make the reason for having MPE
> in the first place go away.

I am hoping that since there is much less graphics involved, which the
emulators you mentioned must handle, that the performance issues won't be as
bad.  Of course, I have no clue.

> Some people have said that the proprietary hardware is
> exactly what we want to get away from, hence an emulator on
> top of an OS that runs on multiple hardware platforms. Again,
> you're sacrificing speed (regardless of the fact that
> processors and memory are getting cheaper and faster) and
> reliability for convenience and portability. While I'm all
> for portability, I wouldn't see that option as a reason to
> keep my enterprise applications on MPE.

Are you still running a 9X7 machine?  HP cannot make money when it's cheaper
to drop off support and coast than it is to upgrade.  It's a different world
out there in the hardware arena and we will have to adapt.

> Why do we like MPE anyway? Stability - OpenMPE's reliability
> and stability would now stand on the underlying OS and the
> emulator's interaction with that OS. Performance - just
> removing MPE from the HP3000 already decreases that, but then
> placing another layer between MPE and the hardware drops it
> even further. Image - from what I understand, that's already
> available on other platforms (Eloquence? I honestly don't know).

There is some stability in having MPE on the 3000.  But there's a different
kind of instability when tied to a particular vendor's hardware.  (Does
11/14 ring a bell?)

> I'm sure an emulator could be created. I'm sure it could
> handle all those things everyone likes about MPE, like file
> equations, build, image, etc. But then what? Advertise it?
> "You loved MPE...NOw get the next to next best thing!" I see
> a FEW people wanting to run it, but not the multitudes that
> some people feel there will be in order to fund and support it.

I agree, that approach won't work.  They will not come play in our sand-box.
We need to invade their turf and show people how reliable computing really
works.  It's the kind of marketing we have always wanted to do.  Why just go
out and continue to preach to the choir.  (In my best John Belushi.  "Are
you with me?  Did we sit back when the German's bombed Pearl Harbor?"  Shhh,
he's on a roll... "Let's go!!!"

> Again, maybe I'm just missing something and need some
> handlholding to be able to understand it all. I don't want to
> sound like I'm against anything. In fact, I'd support
> whatever comes out of all this.

=======================================

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *