In the wake of HP’s end-of-support notice for the platform in five years, we wanted Vance to brief us on the state of the SIB requests and outline the timeline of expected development from the division’s engineers. Vance has been a proponent of Open Source for MPE. In the weeks before the HP announcement, he registered the Web address “OpenMPE.org” for himself, an indication of a personal interest in the OpenMPE movement. It was just another bit of initiative from one of HP’s most productive sources of 3000 engineering. Vance began programming an accounting package for the system as a college student in California in 1978, went into a summer job with HP, and then joined the division even before it was known as CSY. After more than two decades of service to HP and the 3000 community, we wanted his comments and analysis on the Transition, coming from inside the HP CSY labs and speaking as an individual whose entire IT career has been wrapped around the HP 3000. We spoke within a few days of the HP announcement.

It’s got to be a pretty emotional time inside the CSY labs this week. What’s the mood inside the division?

We learned about it earlier, so we’ve had a bit more time to understand it and internalize it. We went through the same shock, ups and downs and grieving, that we’re seeing from some of the customers, expressed on 3000-L. We went through those ourselves in the division. The announcement didn’t surprise us, since we were plugged into the financials and we saw people leaving the platform somewhat regularly. We knew we were on a downhill course, but the timing — I guess you’re not ever quite ready for it.

Did you see something in the ecosystem that seemed to be key, as far as exits and erosion?

I believe the erosion is true, and one reason is we didn’t get any pickup in business after Y2K. We were really hoping for an upswing afterwards, and it just didn’t happen.

What’s the latest word on the enhancements from the SIB development process? Are all the things that were in play still scheduled to happen?

I don’t have the status on every single SIB item. I’ve been more focused on the announcement and getting migration things in place. As far as I know, the LDEV1 greater than 4Gb isn’t finished, and I don’t know if its one-half or three-quarters done. That’s being done in Bangalore, so it’s not as easy for me to find out as the UPS integration project. I just walked down the hall to talk to the guy doing that, and that’s looking good. He’s doing preliminary testing, and we have a SHUTDOWN command that’s complementary to that project, and those are very nearly done. We have to do the administrative part of that, getting it into a patch. The R&D parts of them are very close to being done.

On LDEV1, it might make sense to put that on hold, and see if customers think that’s more important than other activities. We only have a two-year window left here — and that includes the solicitation of what we should work on, investigation of doing the work, doing the work, testing the work, and submitting the work. Just two years, so we need to make sure we have everyone who has time allocated for the SIB process to be able to maximize their contribution. It might be in our users’ best interest to scrap the LDEV1 project — and even though we have some percentage of that work done now, focus on other items that facilitate migration or make the platform sustainable for a longer period. The SIB process we now view as even more important than before the announcement.

Since the SIB is getting more important to HP’s commitments now, it looks like your more public role will get more important, too. Do you enjoy taking a more public position with the user community?

I like being involved with users, and I’ve always enjoyed that part of my job. I really like hearing their suggestions, in terms of implementation ideas or business needs, and being able to deliver a solution to them in a timely manner. With the SIB process I’m able to do that, in addition to scaling it up a bit by influencing what other engineers in CSY work on.

The part I dislike about the role is that I really do like programming, and don’t really like the management aspects of the work. The SIB process has management overhead to it.

You’re one of the more amazing programmers in the community in terms of productivity. How do you balance the personal and professional demands in your life? Stories are out there about things you wrote “because it was a rainy weekend.”

Well, that’s how I like to do it: If it’s a nice weekend I want to go out and do something fun, and if it’s a rainy weekend I have fun programming. Especially if I see leverage, if I’ve already been in the code for something else I’m doing and it’s just a matter of doing a little bit more. I like to be able to leverage as much work as I can into a patch, so there’s less administrative overhead and more time actually creating code.

I’ve sacrificed sleep, to some extent. HP has been really good because I have really flexible hours. They cater to my schedules. I don’t come into the office on Fridays, so they don’t schedule meetings with me on Fridays. They don’t schedule meetings until later in the morning because I drop off kids at school. I teach in one of my kids’ first-grade reading and math classes one morning a week, and I’m involved with a play my kids are putting on, so I do some set-building for that.

You’re one of many MPE engineers at HP who does a lot of work outside the HP campus. Can you describe the lift in productivity you experienced moving outside the cubicle?

Being able to work from home and not doing an hour’s commute each way helps a lot. HP’s just really good about accommodating the crazy life many of us lead right now, being pulled in a lot of directions. I work a lot at night, and my wife’s real busy with her first year of teaching high school. We’ll be up in our office until the wee hours of night, she doing class prep work and me, well, lately, just doing e-mail. I got over 1,100 messages for three days in a row. Normally I’d be doing CSY work at night.

My personal situation is ideal, where I still have an office to come into and have a social life with people at work. I’m not that far away; I just live in the Santa Cruz Mountains, so I can get down here in an hour. I can have aisle-way discussions that are very creative, and at the same time I can just work at home when it’s convenient for me and not do a commute and focus on the few things.

You’ve worked hard on Shared Source, but there’s some criticism that model hasn’t produced much in the way of collaborative results. How do you contrast Open Source with Shared Source?

I think in general people view Shared Source as a failure. I don’t quite see it that way, but I would admit that it hasn’t worked out as well as I was hoping, or the folks at Interex were hoping. I guess we haven’t been offering source code that’s really attractive enough to get people to work on it. We did pre-release FCOPY — it’s not on the Shared Source site, but some other folks have it. The bottom line is that it’s been somewhat of a low priority for Interex, and it’s been a low priority for CSY. It really hasn’t gotten a lot of attention here.

Furthermore, to offer some of the source code that is more interesting and popular becomes very difficult. We have some of the same problems that Open Source would give us. For example the CI, since people wanted the CI to be offered. The CI would involve us including 300-400 files or more, because it includes all sorts of lower level operating system data definition files. The CI isn’t just a program: 90 percent of the CI is in the NL. Customers would have to be able to build new NLs, and link them correctly with the right privileges. It just quickly snowballs into a pretty big problem. That’s why we’ve tried to have the Shared Source self-contained, like EDITOR and QUERY and the Java IMAGE Class Libraries.

Shared Source was a stab to see if some of the low-hanging fruit kind of enhancements that people wanted, in subsystems that HP was really not readily working on, to see if the user community could help. But we were slow in delivering, and HP and Interex both had some issues there. We got stalled with using CVS as a source control tool. A lot of things got dropped, and it wasn’t nearly as successful as it could have been.

Do you think HP’s experience with Shared Source will color HP’s decision to work in an Open Source mode for MPE?

I have not heard anyone use Shared Source as a reason to not pursue Open Source. I think the Open Source dialog that’s discussed around here is more on business case — can we come up with a way that’s good for customers and ISVs and HP? I’ve also had discussions with people offline about whether or not, due to the complex nature of MPE with tens of millions of lines of code, there’s enough enthusiasm, time and talent in our community — a small community compared to others — to really make it work.

Open Source programming is continuing to have more impact on the 3000. Do you think the model has any potential for the HP 3000’s operating system?

There’s a tremendous amount of ramp-up. Most of the code is written in MODCAL, a Pascal-like language, so it’s not in C. There’s a lot of effort HP would have to put into it to get the code out in public. Shared Source had licensing restrictions. Full GPL Open Source wouldn’t have any of those restrictions on it.

At the same time, there might be a model that works better, where we’d target a couple of individual companies and license the source code to them, pieces of source code or whole subsystems like IMAGE. To individuals or to consortiums, so it’s not public for everyone, and those individuals could make a business case, because they have an advantage in that they have source code. If it’s Open Source, it levels the playing field, and it might level it so much that it’s a fairly small pond — and you might have too many fish in one pond for any of them to be able to scratch a living out of it.

Is there enough horsepower in the community, in your early opinion, to take on MPE’s continued development?

The honest answer is that I don’t know. That’s enough to say I have my doubts, but I want to be optimistic. I personally hope there’s enough, but I don’t know. We have people that have been working on this platform for a long time. While they’ve been working on it they’ve gotten married, had kids, some of them are thinking of retirement. I don’t see a lot of fresh blood, young blood coming into the community. There is some, but my perception is that there’s not a lot.

It’s not an operating system that’s taught in universities. I really don’t know if there’s enough interested people with enough time to make it a priority to work for free or whatever business model comes up to sustain MPE, even if they had the source code available.

People have surmised that when HP made an engineering assessment of moving the 3000 to IA-64, the estimate showed is wouldn’t be a profitable position for HP. Is that what happened?

That’s true, but that’s a calculated business answer that sits well with some people. Furthermore, there’s another aspect to that as well. We think that even if we magically had IA-64 here at no cost, the decline in ISVs, a decline we had with everyone expecting IA-64 — if it was there magically, I’m not sure that would reverse any of this. That ties into the business case: if it was there, we’d still see the downward sales figures and the MPE expertise continuing to decline in companies. Jon Backus’ [OpenMPE] group is great, and it would have been wonderful to have that 10 years ago.

I guess nobody was aware back then such an outside group would be necessary for MPE’s survival, right?

I’m sure 10 years ago HP was a lot less open to it, since we had our own training then. We’ve been under-funding the training for MPE for a long time, and I don’t know which came first: whether we just decided here to reduce funding, or we reduced funding that to match the revenue we got from people taking MPE classes. The HP effort that went into the funding of the classes has been on the decline.

Did you see as much of a decline on the tools and utility providers side as on the application side of the ecosystem?

I don’t know if the tool vendors declined less than the application providers. I know that some of the application providers were struggling to make sales on the 3000, in part because of the platform they were running on.

What were you thinking about the day you registered the domain of OpenMPE.org?

I was strongly advocating Open Source back in September and August, internally in HP with some selected folks. I didn’t really have any indicator that HP was leaning toward it one way or another. My managers weren’t saying yes or no. But if we did it, that seemed like the right domain name to have. I got help from Mark Klein and Mark Bixby, because I’d never registered a domain before. We’ll see if it ever gets used or not. I haven’t really given it much thought — I’d forgotten I registered it.

What about Posix smoothing issues for the operating system? Will simplifying ports to MPE help in the Transition, given there’s going to be less added to the 3000 in the face of the announcement?

It makes sense, but to my knowledge we’re not doing anything extra in that type of smoothing. I would suggest that Posix improvements get voted on very quickly by SIGs in the System Improvement Ballot. We will more than likely lump them in with other requests we get over the next two years.

Are you planning on giving the SIGs a real short schedule to get a fresh, revised list of 3000 enhancements to HP?

That’s what would work best for us, to have very short turnaround and not have a lot of items they’re trying to consider, because that just slows things down. Two years, from my point of view, is not a lot of time to get all the things done that I’d like to see worked on. We need to have fewer items than last year, and turn it over really quickly, and have two SIB rounds per year, to accommodate migration options or as people decide to stay on the platform.

Are there non-HP things inside MPE right now that would have to be tidied up or swapped out for the operating system to have a chance at outside development?

Yes, there are. MKS owns the shell. Mentat has the streams that TCP services are built upon. There is an Open Source version of Streams, but we looked at porting that, and it would have been more than a year to get that ported to MPE, so the decision was made somewhat recently not to do it. We renewed our contract with Mentat for the streams code. There’s the potential of some drivers that we share with HP-UX and some of the low-level boot-up code, and there may be some issues there. Possibly all those issues have been resolved because Linux runs on PA-RISC. That’s still an area that would take some investigation.

There are technical hurdles in the sense that we don’t own every line of the source code ourselves. There’s tens of millions of lines of code, so there could be a derogatory comment about a customer or a person in there. Since you don’t know that, we’d probably be advised by our legal folks that we go through the code and sanitize it. If you’re talking about tens of millions of lines of code, that’s a major amount of work.

Could you see limitations that kept HP from going all the way with the 3000 effort, just because it’s a large corporation?

I can share a side that’s not been mentioned. HP as a corporation has given CSY many, many breaks, where we’ve been bailed out by the corporation from profit and financial expectations which HP expects from a mature business like MPE. A startup division here might not have to turn any profit at all, because they’re in startup mode. We’re not in that position. People have claimed we’re a cash cow, which obviously isn’t true now. In a mature market, the corporation has certain expectations for their return on the investments they make in the division. We’ve been exempted from those many, many times. Duane Zitzer and folks before him have recognized the special place MPE has with HP customers.

You’re key to the future of MPE, but you have a career to consider as well. What are your own plans in light of the decision?

I’ve been very happy with the technology and working on the 3000. I’m pretty confident I will be able to contribute to CSY at least until Oct. 31, 2003. Beyond that I don’t know what kind of staffing level there will be in the division, what kind of balance they’ll need to achieve between the Cupertino head count and the Bangalore head count. I don’t have any plans other than those for now. I’m just ready to move forward and get things done for the next two years.