Bob Jolls: Thanks, Pat. After Plan B became Plan A, I was asked to take a new job to manage VS/QUERY and languages at Santa Teresa. Maybe you can't be the outcast - the person in charge of the outplan - and then when they decide you were right, they can't leave you in charge anymore, it's too embarrassing. So I got asked to manage VS/QUERY, which was a project in deep trouble. VS/QUERY was the product that the Marketing Division had said for the last four or five years, "You know, we don't really understand all the things you people out in Santa Teresa are doing, but we want that one. Give us that one." And somehow the VS/QUERY group had managed, year after year, not to be able to deliver a product. And they were kind of caught in a bit of a trap: they would go meet with Moshé and other people back in Yorktown, and decide that QBE was wonderful and they needed to have all the QBE function that could be defined working in release one. And then because they knew SQL was important, they needed to have SQL in there, and they had a few other things in there as well. So they had about forty people, and a mountain of work to get done, and a belief that they had to get it all done for release one. So they would always have reasons - I think at one point, Pat was saying to me during the break - they had a list of problems with System R. You know, here's the reasons we can't get this done; System R has these problems, etc., etc. And when this group started reporting to me, I would ask questions like, "Well, so what happens when you go meet with the people in System R?" "Oh, we don't do that." So they had problems, but they weren't resolving the problems.
Pat Selinger: I think one of them was that every time there would be a disk error, it would stop and say, "YSYSTERR".
Jim Gray: We had seventeen hundred calls to SYSTERR. It was a big barrier.
Pat Selinger: We put Bruce Lindsay on this problem and he went and he counted them and changed them to some other error code. [laughter]
Bob Jolls: And they didn't tell Larry Ellison.
Jim Gray: They all called one routine, called PANIC, and then that routine called ...
Bob Jolls: I'd say there was a sort of a syndrome that was not unlike all the people with the different data models. And I hope I don't sound cynical here, but I'd say that a lot of people got into a syndrome that it was OK not to make progress as long as you had a reason. And of course there were always lots of reasons, so not much progress was made. I'd say that folks on VS/QUERY felt trapped by the fact that they thought they had committed all these things to Marketing, and that Marketing had told them they had to have all these QBE functions, and so on, or they wouldn't "have a product." Finally, after lots of debate about this, and going to lots of meetings about what in QBE couldn't be translated through SQL, and therefore what functions would be missing that Marketing said they had to have, I was frankly confused by this, didn't really understand it. So I finally asked for a list of the ten things that wouldn't be available in release one, because they couldn't be translated through SQL. I then went and met with Marketing, and found out they didn't understand what those things were, either, therefore certainly couldn't demand that they be there. And we managed frankly to scope the thing down to the point where we could get a product out the door. Since I left IBM I don't know exactly how that product has done, but at least when it was released it was a pretty successful product.
Pat Selinger: It's a big money-maker.
Bob Jolls: So they had all the right ingredients for a good product: System R, some ideas from QBE, and perhaps from other things ...
Roger Miller: Bob, QMF is tremendously successful. We had a huge cost base, so we had to have a high price. Customers bought it, so it became very profitable.
Bob Jolls: Who says cost-plus pricing isn't the right way? I guess we had a lot of trials and tribulations over the years with the VS/QUERY group from a System R point of view. Frankly, I think those got ended pretty quickly and they wound up with a successful product.
C. Mohan: So this is different from QMF?
Bob Jolls: No, VS/QUERY was the code name; sorry: QMF.
Pat Selinger: And then we went on to DB2.
Bob Yost: Didn't you actually run that on a bastardized version of System R? For a long while, there was no DB2 to run on. You basically took some code, made it work over in your environment so that you could actually test the query stuff.
Bob Jolls: Right. I can't remember whether they shipped QMF with that black box old version of System R.
Bob Yost: That was just for testing purposes because the underlying system didn't exist at that point.
Bob Jolls: So we're going to have John tell us about DB2? Good.