[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: adoption practices
We agree; my comments were aimed at different arguments,
and thus different people.
> Nick wrote in response to my posting
> I'd agree with Kit in broad terms, but invert that order.
> It can be pretty obvious to StataCorp and many others that
> (1) is true of a package, and apparent that (2) is true,
> but (3), expanded realistically, is the crunch.
> Any idea that (1) and (2) qualify a user-written program
> for adoption by StataCorp misses out what's involved:
> (a) scrutiny of code
> (b) scrutiny of help file
> (c) writing dialog if not done
> (d) writing and running test scripts
> (e) writing manual entry
> (f) tech support
> I did not intend to suggest in any way that (1) and (2) taken
> are by any means sufficient; for an economist, (3) on my list---the
> provision of tech support (including maintenance, enhancements,
> documentation rewrites, etc.) is a MUCH bigger commitment, since its
> cost is the discounted present value of doing that work (and
> competently answering the user questions about the routine)
> from time =
> t to +\infty. Even at a high discount rate, that is a large number,
> and if it is not an affordable sum, Bill G. the economist will not be
> willing to shell it out.
> To say nothing of the costs of Nick's item #4 above: anything
> added to
> Stata has to "play nicely" with everything that is already there, and
> that requires (using Stata's published standards; see Bill G.'s SJ
> article on software certification) a LOT of testing (and retesting,
> every time a single new feature or bug fix is made! On every platform
> supporting Stata!) Yes, if I was on the StataCorp payroll, I
> would not
> be real keen on adoption, as much as I would want the package to do
> everything for everybody.
> StataCorp have, IMHO, taken an alternative strategy (to creating a
> bloated behemoth with manuals no one can lift): they have made it
> trivial for you, me, Nick, and everyone else to set up our favorite
> procedures and distribute them costlessly via SSC, user
> sites, SJ, etc.
> Do I have the same confidence that any user code (including
> my own) is
> as bulletproof as what comes out in 'update ado'? Of course not; the
> quality of the offfcial updates is what I'm paying good money
> for when
> I buy a new version of Stata (and those who are not eager to update
> ought to think about that). But just as in literature, some
> authors are
> better craftsmen/women than others...
> Kit Baum, Boston College Economics
* For searches and help try: