It is the present aim of the R project to produce a free implementation of something "close to" version 3 of the S language and to provide ongoing support and maintenance for the resulting software. Some members of the R core have proposed that future developments in S version 4 should be also tracked. At this point it is unclear whether this will happen.
One development which would help R a good deal would be the development of an integrated graphical user interface. Some initial work has begun on this and I believe that it is something which will come quite quickly.
My personal future interest in R is mainly as a user. Given the investment I have made in it, I hope that I will be able to get substantial use out of R for statistical work and teaching.
Working on R has shown me that there a number of interesting questions related to building statistical software. My own conclusion has been that it is important to pursue efficiency issues, and in particular, speed.
As noted in section *, Luke Tierney performed some experiments with R to see what kind of speed increase could be obtained using byte-code compilation; the indications were that a speedup by a factor of 20 might be possible for some computations.
There is other evidence that a factor of 100 (roughly the speed of unoptimised C) might be possible with compilation to native machine code. With this level of performance, there would be no need for any foreign function interface and all computations could take place in a single language environment.
I am intrigued by what such an environment might offer. An increase in performance of this magnitude is likely to produce a qualitative change in the use it gets puts to.
The difficulty is that the creation of such a compiled environment requires the hand of an expert in compilation. There is a real problem in finding such an expert who is also aware of the type of problems which statisticians handle.