RB is a nice product. And, with the addition of the OpenGLSurface control (FINALLY!), it's getting better. Certainly, if you need real OO in your project, RB will beat Rev. Similarly, if you are doing heavy number crunching, RB will easily destroy Rev (since numbers are strings in HyperTalk). Then again, based on the number crunching, you may be better off using C, Lisp, APL, J, or a myriad of other languages over RB

Those points aside, I feel it necessary to really identify the major difference between Rev and every other development environment out there. Rev is 100% live programming. I can't stress enough not only how cool that is, but how much more productive it is.
Programmers who swear by (example) Python over C++ aren't excited over built-in hash tables. They are dancing because they can iterate very fast with an REPL (read-eval-print-loop). Rev takes that paradigm and blows it out of the water. There's no such thing as having to rerun your program from scratch, reload the data, and retest. You just keep tweaking the same, always running application, until it works. You have become a the sculptor of a solution instead of a programmer. And once you experience this, you won't want to go back.
Bottom line: there are certain applications for which Rev isn't well suited compared to other languages. I wouldn't want to do finite element analysis in Rev

But, when the app I'm working on fits in Rev's domain, I'd rather not work in anything else due to the workflow of the environment.
Jeff M.
P.S.
I do have to go on record against the 2 biggest features of 3.5 (behaviors and data grids). The concepts of both are wonderful, and should be awesome features. But behaviors having to be buttons is a joke, and wreaks of "HACK" to get a bullet point feature in. And the data grids, while a noble effort, are just a side project that don't even come close to being truly integrated in with the rest of the environment or intuitive in any way (like the rest of Rev).
3.5, IMO, was a step backwards by the Rev team. 4.0 looks to be a giant leap forwards, but I would kill for behaviors and data grids to be implemented "properly"
