LateNightHacking Louis Journal

Auth

I was thinking about the instrument service a little bit. One big component of a high quality isntrument service is a robust distrubuted object model. We've got some code which is a step in that direction. However, I think to do it really well we would also need a good inference engine or expert system to analyze the crappy incoming data. And of course a very flexible data model and data warehouse to store and index the incoming data. I was also thinking that adding another layer of indirection might make the whole thing a bit more flexible - instead of saying "give me the ticker for this instrument", you could say "give me what Redi wants for FIX field x for this instrument". Then you could drive the whole thing with configuration metadata. (Pragmatic Programmer tip #38 - Put abstractions in code, details in metadata.) Then you just have to manage that data.