Stata seems to now have all the parts (graphics language,
programmable dialogs, ado language, Mata language) that statisticians
could want (what by the way is missing in terms of tools?).
Development of Stata can proceed along the lines of additional
solutions or speeding up existing solutions, by users of Stata and
Stata Corp. Stata 9 came with enough additional solutions and GUI
changes to warrant an upgrade, on top of what must have been a very
substantial effort by Stata Corp to produce Mata. I remember reading
somewhere that Stata Corp viewed itself as knowing less statistics
than the users, and it was a goal of theirs to provide tools &
solutions. Along these lines, Stata is a very impressive application.
I'll still use Data Desk for certain EDA tasks and R for others
(Bioconductor.org), but I still prefer Stata from a non-
statistician / user point of view.
The situation is just like that with any other
programming sub-language of Stata.
Perhaps 100 people write Stata programs
seriously and also put them in the
public domain; no doubt rather more people
write Stata programs, often but not always
more specific, and keep quiet about it.
Many more users never write programs at
all, but they still in _all_ cases
gain from the programmability of Stata.
Even if you never use user-written Stata
programs, you can't use Stata seriously
without making heavy use of Stata programs.
Mata will be the same. You needn't use
it directly and deliberately to benefit from it.
The leverage it gives to Stata's developers
and to user-programmers will benefit all.