The "Heartbreak Hill" of the Masters Program: the Thesis arrives
Well, I've narrowed it down to three topics to choose from for my masters thesis.

The first one offered would sound really good to the folks at Progress (if I cared about that, which I'm not sure I do.) It's about how to replicate a running database. Professor Friendly (not his real name :) ) has some theories along these lines about dividing up queries by matching them against certain templates and using these to allocate the queries among partial replicas of the database (subsetted by table) which can be generated or decommissioned dynamically based on load. I see this as possibly technically doable but with some pretty dramatic drawbacks in terms of real world use -- which means that the supposed applicability to Progress is actually a mirage anyway.

Professor Theory (see note about Professor Friendly) has offered me a straight implementation problem. He (or a friend of his, anyway) has developed and proven on paper a parallel algorithm suitable for reducing the complexity of state models so that the models can be more easily used to prove stuff (see the joke about the difference between a physicist and a mathematician.) He wants somebody to develop a running version of it. A mind-stretching problem (even fully understanding the problem and the solution is going to be some work for non-mathematical me) but one with little direct or visible applicability to much outside the theoretical world.

And finally, just today I managed to talk with Professor Busy's (ibid., op.cit., etc.) TA's research assistant's scientific programmer... who has outlined a need for a peer-to-peer dynamically scaling world state repository suitable for a bunch of evolutionary agents to run around in. I went to this professor originally because I was interested in evolutionary computing, but this project would touch upon absolutely no evolutionary concepts... but the dynamic scaling, load-balancing, neighbor-interactive nature of the project is also appealing. Also appealing is the fact that I do understand the problem and it's not databases. And that there's a reasonable chance it would involve (or at least let me rationalize) a trip to Edinburgh, Scotland, which has always been on my list of places to see.

I am leaning toward the last project -- I am tired of databases and not convinced that the templates/replication idea will amount to anything useful, and I'm not sure there's a whole lot of reward in understanding a mathematical project that really can't be used anywhere else. The P2P world state thing has its drawbacks too -- poorly (read: not) spec'd, some political shenanigans associated with this multi-university project, and I'd have to work a little harder to get it approved because technically the project is outside the PDCS department. But I actually feel some excitement about it, unlike the first two, which is not only a clue but important in its own right, right?