First of all is it something that one
person can do in a reasonable
time-frame?
Most likely not... ThinkOrSwim has been in the business for a LONG time, so duplicating all of their effort is pretty much impossible to do with one person. If you start small you might be able to do some things, but it's quite challenging. Furthermore, if you try to build ThinOrSwim in Silverlight it's almost like building Windows in JavaScript: it simply doesn't make sense.
Ok next question is how would the
performance of the Silverlight be
running such a memory intensive
application? Will it scale?
TOS is not just a web site, it's a brokerage... that means that they have to maintain connectivity to exchanges which is VERY bandwidth intensive and hardware intensive. You would have to connect to exchanges too or, at the very least, connect to another brokerage. In other words, your Silverlight will only be a front-end and nothing else. At some point you will have a central server where your users connect to unless you're just building a front end to another brokerage API like the Interactive Brokers API
Nearly all trading platforms run
locally on the machine, there must be
a reason for that instead of building
it on top of Adobe Air or Silverlight?
Trading platforms tend to be very CPU intensive too, so the closer to the hardware the better... Adobe Air and Silverlight add a level of indirection which impacts performance.
Last but not least, would users be
frustrated by forcing them to install
Silverlight plugin or is it something
that people have no problem installing
on their machines?
I don't think that would be a concern... installing Silverlight is pretty easy to do, so I doubt they will see it as a big pain. The only issue, as far as I can see, is with the performance.