[R] A long digression on packages

[Berton Gunter]

$.02 (no more):

1. R and its packages are big and diverse, growing rapidly. No simple phrase
to describe this, but perhaps " the most important contribution to data
analysis and statistical practice since Fisher" comes close.

2. Ergo lots of diverse information, with little or no way to classify and
organize. But this is THE BIG THING in IT today, is it not? -- witness
Google, Yahoo search, etc. Everyone says understanding and technology to do
this well is in its infancy. So that R and its community struggles too is no
surprise.

3. As desirable as efficiency and redundancy reduction is, R's nature and
design mitigates against it: core R is centrally controlled (of course!),
but the point is **NOT** to restrict contributed packages (core R is the
engine and coherent point of entry into all those packages). So the best we
can hope for is that package contributors will do their homework before
writing to see whether their intended functionality is already there or
could be just added to someone else's. Obviously very subjective -- and
could be difficult due to terminology variation (different disciplines that
use different terms to describe the same statistical functionality).

4. So other than package writers putting as many keywords as they can into
their packages for search engines to hit -- and perhaps some limited
organization by dedicated workers to organize/bring together "obvious" stuff
like graphics or econometrics or geostatistics, etc. -- it seems that all we
can reasonably hope for is to integrate the search functionality into core R
and R directory structures. But wait! -- this is exactly what the R team is
already doing.

Cheers,
Bert