Waiting for the Key to MPE’s New Factory

Mark Klein has masterful MPE experience, and he wants to put it to work on behalf of homesteading customers. The former R&D director for ORBiT Software returned to his independent development and consulting roots during the first year after HP’s decision to exit the 3000 market. During most of that year Klein has served on the board of directors of OpenMPE, Inc., the organization dedicated to extending MPE’s lifespan beyond HP’s plans for the 3000 operating environment.

Few periods have been as quiet as the last three months of OpenMPE’s existence. The organization has been conducting careful, intensive talks with HP to tackle the MPE licensing issue, a subject that is essential to the existence of any 3000 hardware emulators. Those emulators can make up the next generation of HP 3000 hardware, once Klein and the other board members wrest the details from HP on new MPE licensing.

New MPE licenses will open the door to OpenMPE’s most creative work: marshalling the MPE technical expertise that has become widely available into some sort of virtual lab. Most homesteading advocates understand that the 3000 must continue to grow, and so OpenMPE must demonstrate to HP that it’s possible to maintain MPE outside of HP lab.

Though he’s not campaigning for the job, Klein is a good candidate to head up such a lab. Leveraging GCC work from the University of Utah on HP-UX, Klein did the bootstrap porting of the GNU Software Foundation’s C++ compiler to MPE, work that set the stage for the HP 3000’s compatibility with other systems. Without a C++ compiler, things like Web servers, Java, Samba file sharing, Internet domain name services and many of the platform’s Posix tools wouldn’t have arrived on the HP 3000 in time, if at all. Working nights and weekends for the equivalent of a year of full-time coding, Klein brought this new language to the HP 3000, one that opened the door for the HP 3000’s interoperability.

Between his experience developing for HP — which honored him with the HP 3000 Contributor award — and years spent managing ORBiT Software’s developers in an era when 20-plus years of MPE background was the norm for that group, Klein looks to us like a smart pick for an OpenMPE post that could emerge this year. We spoke with him about the prospects of making MPE thrive outside HP, first in December and then again in late January and February — the months when the air grew thick with anticipation over HP’s MPE homestead licensing arrangement.

Can MPE be maintained by experts outside of HP?
I think that there is a very good possibility that it can. There are some very good people with an in depth knowledge of MPE internals, Networking internals, TurboIMAGE internals, etc. that could make this work. If anything, these people could be the “architects” for a future development effort, leveraging their skills with others that don’t have that specialized knowledge.

Is there enough non-HP expertise available around the world to enhance MPE with new features in the years to come?
The same answer I just gave applies here as well. The main point is that I think there are enough of the architect quality people available to steer such an effort. Without that, it is my opinion that the chances of maintaining, let alone enhancing, MPE suffer significantly.

Significant parts of MPE have been from vendors other than HP. How do you feel about the prospects of replacing these sections with Open Source alternatives?
I don’t see any reason that it couldn’t be done with the exception of some of the lower level IPC code. Whether or not it is realistic to do this is another issue. I also suspect that there are other issues with the Open Source licenses that might make this a difficult course to follow.

One such example is the GNU requirement of “making the source available” to anyone that requests it. It is my understanding that HP desires not to make the MPE source openly available at this point in time, so incorporating GNU licensed code might not work. On the other hand, I can also see a scenario whereby the GNU code could be “plug replaceable” in the field by the ultimate user and thereby not cause MPE to fall under the GNU license. I’m not an attorney, so I could be all washed up here.

Has HP surprised you in any way regarding its willingness to help MPE survive beyond 2006?
Not really. I think they are trying to do the right thing for their customers as they’ve said all along, even if we don’t agree with their conclusions. It is possible that the “business case” for MPE beyond 2006 was not well enough presented before last November for them to initially consider it. Since that time, it has become obvious that there is a case.

Why won’t a traditional Open Source program work for continuing with MPE development?
This is a tough question, and a lot of people will disagree with me on this.
If you set the intellectual property issues aside, I don’t think that the critical mass of developers is there to make it work. With a small pool of developers available, unless there is focus, there will be no movement to benefit the majority of MPE users. Would you rather have 20 different developers each releasing 20 different versions of MPE, or would you rather have 20 different developers working together on one release?

Let me use my port of GNU Compiler Collection [GCC] to MPE as an example. I’m one of a large number of contributors to the GCC effort. Because I felt a need, I did the port. But, while I’ve contributed the port back to the development effort, it has never been fully incorporated into the sources. Why? Because that port has a very small attraction to the masses of GCC users. The core development team wants to do things that benefit the masses.

Now, reduce that by an order of magnitude or two for MPE. You get the 20-release scenario. I think a focused development effort is the only way in our universe that we will see progress.

Can an independent effort to maintain MPE succeed if IMAGE is maintained by a separate company? Does one organization need all the code for the effort to be effective?
I think it can succeed. Heck, when I got my first consulting arrangement with HP in the early 1980’s it was for the database lab. That lab was completely separate from the MPE lab at the time. While there might need to be close partnership or relationship between the companies to make it work, I don’t see why it couldn’t.

You’ve said that HP seems to be willing to license MPE source to individuals. What’s the benefit of that kind of development plan, versus the Open Source arrangements that have powered Linux?
I hope I am not reading too much between the lines, but I have not heard them say “absolutely no” to the source licensing issue, and I would be interested in being such a licensee. Earlier I brushed aside the intellectual property concerns. I think those are major concerns. Private licensing can address those issues to HP’s satisfaction, I would imagine.

Some companies want to simulate MPE on other platforms today. What’s the reason a customer might want to do that, versus the harm they might do in choosing a simulated solution?
I cannot say that there would be harm in choosing a simulation method vs. an emulation method. The major difference between them is that you need to be able to recompile your software and possibly change some of your processes to work in the simulated environment. In the emulated environment, you should be able to pick up your existing applications and data, lock, stock and barrel, and move them to another platform. There are arguments that can be made on both side of this issue. I’ll bet that if emulation becomes a reality, we’ll see a presentation called “Simulation or Emulation” at some point in the future. I’ll leave that to the marketing gurus of the interested parties.

The only compelling technical reason for emulation would be needing complete binary compatibility, and you can’t find the source code for some application you absolutely require for your business. In that case, doing emulation is the only course, short of re-inventing the application.

Now that HP has agreed to license MPE to new users on a hardware emulator, what’s the next biggest roadblock to making homesteading a viable Transition solution?
I think total cost of ownership and total cost of development. There are discussions about the hypothetical cost of a MPE license on the OpenMPE list right now. There have also been some discussions about what the emulator might cost. Sooner or later it will come down to dollars. The emulator companies will need enough return on their investment to justify that investment. The homesteaders will need to receive enough value to justify spending those dollars.

What do you see as the real deadline for a customer to leave their HP 3000?
Support seems to be available at least to 2010, and the hardware availability might well be answered by an emulator. What does a customer have to consider when deciding their migration timeline?

I don’t think there is a real deadline, as much as I hate to admit it. By that, I mean that each company will need to decide this themselves, based on their own business conditions. I know firsthand of one company that has decided that migration or replacement is just too expensive for them and that they will stay with the 3000 until they can no longer use it to run their business and then they will close up shop. I wonder how many others that we are not hearing about have come to this same conclusion?

So what’s going to motivate change in the 3000 community if there is no real deadline?
Politically, companies are going to be faced with a migration, even though there is no deadline, by virtue of the fact that HP has made the decision to end the platform. Look at what QSS is doing [by porting to Linux]. They’re making the move, even though there’s the possibility of staying on the 3000 for another five to 10 years. HP is saying do something. That’s enough for a company doing business-critical applications on the 3000.

Is there a need for an in-person training initiative to foster 3000 and MPE skills? It seems like so much of the training now is for migration tasks. Does the market need more people that know how to manage the 3000?
That’s a good question that I really am not sure I can answer since I have very little to do with the operational end of the 3000. If you listen to some people on the MPE specific email lists, there is the thought that MPE can grow, post EOL. If that comes to pass, then surely more training will be needed. I guess this could become a chicken and egg thing, though.

How much of a lift does HP’s decision to license MPE for emulators put into the OpenMPE movement?
Well, you’re asking me, so you’ll get my opinion. At the formation of the OpenMPE group, I think I was the lone voice talking about licensing issues. Most everyone else was concerned about Open Source. Slowly, I think it became obvious that the number one issue really was licensing. Without it, nothing else could happen.

With a replacement license scenario, the MPE universe would slowly get smaller until it completely died. With a licensing scenario whereby new MPE licenses could be created, the universe could grow.

I’m ecstatic about the decision to allow the creation of new MPE licenses. I can not express how big I think this really is and how favorable it can be to the people that don’t want to see MPE die.

Your expertise is unique in the 3000 community. Is there enough hope on the homesteading table to keep you involved with HP 3000s and MPE?
I have always said that when I retire, it will be from the 3000. I’m too young to retire for at least another 15-20 years!

Is it important for the OpenMPE movement to have its own lab?
I don’t know that it’s necessarily important. A lot of the stuff could potentially be done by the emulator companies. However, there’s a handful of us who think that MPE could be supported beyond end of life by others than HP, in case an emulator company doesn’t come along and step up. When you start to look at people of the caliber of those leaving CSY now [through layoffs], there’s a fear that the capability to make critical bug fixes won’t exist in two years, or even after HP’s end of life deadline.

Based on what Dave Wilde said at HP World, they’re going to support a homesteading movement. I don’t think they had their ducks all lined up in a row when they made that statement in terms of how they’re going to support that movement, if they lose all of their MPE knowledge base.

One of the things that OpenMPE is attempting to do as an advocacy group is making certain that kind of knowledge exists outside of HP. I think that’s the major goal right now.

Can you get to the point of being able to pick up some of the departing MPE talent through contracts? How do the expectations from the homesteading community align with the process that’s in front of OpenMPE?
People tend to want everything, and they want it in backwards order. It’s the chicken and egg thing: where do you get the funding for [those contracts] if there’s no possibility of you doing anything with the people you’d get with that funding?

There’s a logical progression you have to follow. Let’s come up with intermediate tasks that will get you to the greater goal. The most immediate need is to come up with a licensing arrangement. If you can’t license MPE, everything else is moot, in my own opinion.

Some customers still wish HP will change its mind about its 3000 business. Does it make any sense to you?
No. Whatever damage has been done to the market has already been done. As a die-hard MPE-er, I would love to see them reverse direction. In my opinion, I think there’s too much momentum going downhill away from the platform for them to do that.

What about changing the end of support date?
That may be something worth considering. 2002 has been a planning year, and migrations haven’t even begun yet. HP made the announcement so late in 2001 that the budgets for 2002 were already in place [for customers]. Very little activity could happen in 2002. As a consultant I sat on my butt for much of that year, because none of the migrations that were forecast for 2002 happened.

We’re more than 14 months into this before the larger companies are starting to move. Can they move effectively and get off the platform before the end of support [in December of 2006]? No. So there’s a strong argument for taking that end of support date and adapting it. It could be possible on a case by case basis.