Richard Walters:

"M-Boerse" was represented by Dr. Kirsten, now chairman of MUG Germany, and Wilhelm Pastoors, one of its coeditors.
As we think that what Dr. Walters said in the interview may be of interest in the current discussion we decided to publish the interview on the net now.

Dr. Walters:   At this point, given the announcement we heard an hour ago, I can't give you a direct answer. I can speculate, though: there are two directions that the M-community might go. One is to say we have only one commercial vendor, that commercial vendor does not support the MDC and the statement that they published about the acquisition of Micronetics doesn't even mention the MDC.
They feel that MDC moves too slowly for the commercial market, as they stated in their document.

So, there are two things that might happen: one is that MDC will die; the other is that the M community will say we can't let that happen, since M cannot evolve with the name Cache' replacing M and becoming a commercial product, not an ISO standard.

There are many computer scientists who have worked with the M language who feel that it is a major contribution and would like to stay a part of that evolutionary process. If the M Community decides to try to preserve the M standard, there are again, two options.

The first is to support the remaining or help develop potentially new commercial versions; the second is to support a public domain effort that might lead to several versions of M and its global handler. These two options are not mutually exclusive; in fact, they could easily be addressed in parallel.

Let's consider the second option first.
Some years back, a small group formed a distributed partnership that we called Generic Universal MUMPS, or GUM for short.
There were about four or five contributors, including LD Landis, a former student of mine (Terris Linenbach), Steve Morris, Dr. Cui Zhang (a Computer Science professor at Sacramento State University), myself and others. GUM has been led mainly by LD Landis, but it has been housed on a UC Davis ftp site (and will soon be on another web server at UC Davis). This project got as far as completing a working version of a "Global Handler," that is to say, an implementation of the M global structure in such a way that data could be stored in a MUMPS structure, with the functionality of creating and modifying globals and manipulating them via $ORDER, $DATA, $QUERY, and $GET. The GUM team intended also to implement an M compiler so that all language elements, not just the database, would be available in a public domain version.

GUM has not yet achieved its goal of implementing a compiler, but the global handler has gone through some improvements, a process that is continuing.
With the Intersystems announcement, interest in GUM has heightened considerably, and the M community needs to ask itself: is this a viable way for us to proceed today?

I don't know the history of the PERL language but it would be an interesting one to explore because PERL came out of nowhere. Five years ago there was no PERL. Now there is PERL on almost every machine. There is even a user's group meeting scheduled this Fall. How did that happen, what could we learn in the M community that might make us want to try to apply this model in the evolution of M?
I have set up the web page for GUM code to be downloaded from a server at UC Davis. I offer this to the M community as a way to provide software using freeware models like GNU or BSD (details remain to be worked out). I think that such an effort is worthy of support, and I hope that people in the M community will join in seeking grant support for people to develop versions of M to place in the public domain.

M-boerse:   You stated that the global handler is too slow.

DW:   No, what I said was that earlier M implementations at UC Davis were too slow to serve as starting points for a new M compiler. We had several versions of MicroMUMPS running on a good many different machines, but their performance was much slower than existing M implementations of their time, and the code was written using a technique that was cumbersome. Another implementation of M by a student at UC Davis generated P code which was then interpreted. This version also proved to be orders of magnitude too slow to serve as a base for the present effort. Actually, the GUM global handler is quite fast, certainly fast enough for a public domain version.

M-boerse:   What is a global handler?

DW:   In M, the term "global" refers to what in other languages is referred to as persistent data. Global variables are stored after a program that created them halts, and they can be shared by multiple users. (M "local variables" are the equivalent of global variables in other languages, in that they are available to all procedures running in a workspace.) The GUM global handler creates a B-Tree structure of the hierarchical M globals, storing them using the same "canonic" sorting, based on numerics, then ASCII code values to sort string subscripts. This structure outperforms relational model joins, as has been shown in many studies.
GUM allows globals to be created and modified, and it also allows them to be retrieved using the functionality of standard M functions such as $ORDER, $DATA, and $QUERY. With these functions, I can work with M-type data, but I don't need other M commands (besides the equivalent of SET and KILL).
I can use a C call to $ORDER and go out and get my data - implicitly persistent data - from a C call that goes to the global handler. So in the M language the concept of persistent data is independent of the concept of the interpreter.

When I say that we have a global handler I say that we have a database that is fully functional but I don't have the MUMPS commands, operators, and other features that are part of the language.

M-boerse:   A part of the Linux project was that a lot of universities directed their students to make certain parts of the system. So from different universities came different parts that contributed to the whole project. Now, the Frankfurt university hospital supports Mumps, to give some names: Prof. Giere, Dr. Kirsten - could you imagine that there will be a concerted action by - for example - the department of informatics at the Frankfurt University and UC Davis and other universities ... to get some more support for the realisation of the idea of a free mumps?

DW:   I am not sure - I think that today's announcement will act as catalyst for actions of this type, but probably involving a much broader segment of the M community. One of things what Linux has in its favor is that Linux is Unix based. Unix is taught in computer science departments around the world. Historically, when Bell Telephone first developed Unix, they sold it for 20.000 $ to businesses as an operating system . They sold a license to the entire University of California for $150 for unlimited licenses. In other words: they gave it away to the educational world saying "it is yours, you can go develop it, you can work with it however you want, here is the source code, have fun!" BSD is one outgrowth of that effort. As a result, UNIX became the principal operating system taught and used in computer science departments around the world, and with time it gained a larger share of the commercial marketplace as well.
MUMPS, on the other hand, ran on dedicated computers with a proprietary operating system, and computer science departments were not willing to dedicate an entire computer just to M. As a result, there has been little interest in the M language in computer science departments. A few individuals, such as Kevin O'Kane (then at the University of Tennessee) and my group at UC Davis developed public domain versions of M, but there were very few.

M-boerse:   So, we have to see that the computer scientist base for developing an M Database is much smaller than it was for Linux. We have to go by personal contacts. The situation is that Frankfurt supports Mumps very strong and there may be a chance to make personal contacts to the department of informatics to show what's going and maybe there is one professor who says: "ok, I will go to support it, I will direct my students to that project." We have to build a net of people.

DW:   ... Yes, I think that the Goethe University has a very good potential to serve as such a center. I hope we can interest others in this effort as well. The GUM project started as a distributed effort, and even if I retire in a couple of years, it should be able to continue without me.

M-boerse:   But you are the person identified with the GUM project.
Who is the owner of GUM?

DW:   The University of California is because of me. Not because I did it but because I got people to get started on it. It has a life of its own. I also think that M is important and I don't want to see it die. I'm willing to put an administrative effort into it, and it is a working model at present. But I cannot contribute to the actual compiler development. That would require help from groups and institutions such as Goethe University and the growing members of the GUM team.

M-boerse:   At the moment situation is that there are only a few active members of the group. If we could find a few more to join, it would significantly improve the situation, right?

DW:   All of these people have their other jobs, I don't think they have had time during the past two years to work on the compiler. It takes an extra effort , an extra incentive to say: "right, instead of going home and playing with my children I'm going to stay for an extra 5 hours and work on the compiler". That's something that may happen, we may get people, especially after today's announcement.

M-boerse:   Just to get an imagination: how much work has to be done?
There is a lot of overhead created when a software product is manufactured for the commercial market. Sales Representatives, technicans who install and support, consult. Just to break it down : how many people do you think work at Micronetics or Intersystems departments of development?

DW:   I truly don't know. Not a large percentage of the total group in any one company. I have heard that Micronetics had two lead developers and about 18 others working on various versions. DSM, so I understand, had about seven working on the system, and DTM perhaps a few less.

I think we might be able to get enough people to develop one or more public domain compilers. With the incentive we now have, many people are saying: we can't let this happen, we can't let Mumps die, I think we can have a version that was operational for the MDC to start to work with in less than a year. And that assumes that we have a leader who knows something about it, and who can assign to other people different parts of the parsing, the lexical analysis and so on, I think that could be done. It looks as though we do have such a leader willing to head the team (LD Landis), so we are off to a good start.