We started, back in 1992 with an approach to IBM's main executives charged with promulgating OS/2, and in particular making it compete with Windows.

John Soyring still works for IBM. It is a moot point whether or not OS/2 does ;}

We spent an unprofitable hour on the phone with him and some IBM folks on 20th November, 1991.We assume they simply didnt get it , since we never heard another word.

The clock ticks forwards to 1998.

Java has now been "invented". OS/2 is almost history. Meantime a smart IBMer named Mike Cowlishaw has been busy. He apparently realised two things:

A. That it would be possible to make a Java pcode implementation of Rexx, IBM's batch control language by writing a compiler. Indeed one of the points we made to Soyring was that heterogeneity is a feature of the Code Server approach. It is a great way to support multiple platforms, with or without the use of portable (processor- and OS-independent) pcode. 

B. That the Java Virtual Machine (JVM) is the key to portability across networks, rather than the Java Language, which is nothing more than a slightly messe up version of C or C++. 

Imagine our delight at this, of course this whole project started because of REC's experience and expertise in writing portable software and pcode interpreters, with specific reference to networks.

So Mike seems just the guy. 

What happened next was kind of funny. We approached him (he works at Hursley Park, IBM's technology centre in the UK, and the home of much of OS/2), and he referred us to an executive named Simon Phipps. Simon appeared to be interested but kept referring to "the folks in Austin". So we contacted John Soyring (and just for good measure some of the folks he referred us to) again.

Eventually REC Software Inc. gets a call from Dun and Bradstreet Canada (!!!!!!) After some discreet and slightly outraged enquires it is determined that D&B were set on the trail by an attorney in IBM's Toronto offices. When the smoke clears and IBM and D&B both apologised to us, IBM finally gets its act together and decides to let us talk to Frank Casey, who is based in New York State, close to the heart of the "real" IBM. He in turn sends us back to Hursley Park, for an interview with a very opinionated techie named Alan Webb who apparently made up his mind that our idea was such ancient history that IBM would have no further truck with it. 

And that appears to be the end of it. Casey told us that IBM sent out some sort of proposal, or request for feedback - we have no idea what, since we didnt see it. At this point we believe that IBM was merely going through the motions. 

Alan Webb's main point seems to have been that dynamic linking does not necessarily have to be married to association (ie the finding of all components), the same point made by Richard Reisman (see under BTG). On the contrary, we have yet to see a system that could escape the requirement of namespace resolution for named modules and entrypoints before execution can begin, unless it is in some way restricted (ie made less universal) by pre-classification or "slot allocation". Apparently there may be an operating system called SPRITE which may relate to our idea, but he wasn't interested enough to give us details. Aha! (29/4/99) It looks as though the reason we couldnt find references to "SPRITE" is because it doesn't exist - he must have meant to say "SPRING". Spring is Sun's "research Kernel dating back to 1991". Hmmmm.