>> Patching a laser in instead of the regular diode shouldn't be terribly
>> difficult either.
>I also suspect that that's be harder then you think. The trick is to get
>it fast. The commercial products don't use powerful lasers, they use damn
>good optics and signal processing.
Microwave is just slightly less picky then light, but not by much! ;-) It is very possible to connect a laser (and some support circuitry) in replace of a low powered IR emitter. Sure, it would probably require the use of an optocoupler, a driver circuit for the laser pointer (partly to keep it from burning up), and some random optics. And yes, laser pointers can modulate very fast. Some have little regulators built in -- this can be as elaberate as a CdS cell in a feedback monitoring fashion or as stupid as a resistor. The resistor trick is becoming more and more common as laser pointer production has reached an all new level of cheapness. (Bad for laser pointer life, good for the laser pen h4x0r) Anyway, these little regulators will limit the frequency greatly -- they need to be taken out if possible. The receiver will have to be a bit different, unless you decide on using an IR laser. If its a photodiode, replace it with one of similar specs in the visable light range. This might be even tricker then the emitter portion. Who knows what the sensitivity of IrDA equipment is. There could be a very elaberate high-gain transimpendence amplifier on it, or it could be similar to your VCR's IR input.
I have read about a lot of hacks that amateur operators have done with lasers. Its almost comical what can be done with various household items (more specifically, household 'items' of a geek).
And you are right, optics is where its at. You need good band-pass filters, a lens system that can pull as much light in as possible while tightening in on a distant point, and a sensitive receiver. It may sound funny at first, but a long tubular $20 telescope at toys'r'us can do the job.
I really think that someone with some spare time, basic high school electronics knoweledge, and the right equipment could create a very usable link. Who knows how far it would work, but i dont think it is as impossible as it sounds. Without counting the laptops with IrDA support (and the possibility of destroying the IrDA port ;-), i can see such a project costing less then $100 easy.