Thanks for this link - forgot about that project. Course, that one is trying to do a 50 year prediction. Even then, it only took them 3 months to do 880K modeled years. Since a hurricane model is only really concerned with at most a 5 day prediction (more if time allows, but errors make unusable), a similar program would be able to crank out a model analysis in no time whatsoever....

Sorry for being "wordy" when I mentioned it first time. I gotta remember this is a meteorology fan board, not a computer science one. ;-)

Oh, and as far as the security of the system, I would think it's negligible. First, somebody would have to want to hack the client to see how it crunches the numbers the NOAA sends it. Second, they'd want to bother to try and falsify the analysis their computer did. Third, it's unlikely that any falsification of the data would really change the overall model that much. True, there's some "butterfly effect" stuff going on (you know, where they say a butterfly flapping its wings over Kansas causes a typhoon someday later in the Pacific), but we're talking about thousands of data points, dropsondes, etc, and I don't think it would matter much. Plus, you call the thing experimental and leave it at that. Not like some of the loopy models this year that take storms to South America are all that useful, either. :-)

It's a shame this isn't being done, or at least, not openly publicized. I'd love to work on something like that. Actually make use of my BA in Mathematics and BS in Computer Science, instead of [shudder] teaching. :-) (Just kidding, I'd still teach, best job in the world, even if the pay is miserable and there's no respect from the students...)