Re: Data does NOT belong in Debian (was: Stop Archive bloat)

[eichin@thok.org (Mark W. Eichin)]

I also find it *very* useful to have doc-rfc and doc-iana.  I'd keep
fortune-data out of tradition.  On the other hand, I'd ditch all of
the linux magazines (lg, pluto, I think there are others?) without a
second thought...

I'd argue that doc-rfc has sort of the same niche as doc-HOWTO.  Not
sure if I can define that niche, though, other than "developer-useful
datasets"; the coastline data, like the tiger/line map data (5 cd's
bzipped, mmmm) is domain-specific, but not
{debian,linux}-developer-specific... and debian has always been de
facto "of, by, and for" developers.

I like the idea of a packages-style installer for these things,
though; if we go way back (pre-debian, pre-redhat) the BOGUS release
(which evolved into RedHat, sort of) was a lot like the *BSD Ports
system, in that you could have a URL and an md5sum in the config
file... Ports has probably advanced, you'd really want
	* set of mirrors (multiple or pattern URLs)
	* md5sum for exact match, but easy-to-upgrade option

The problem is that if a data set evolves independently of/faster than
our release cycle, a package-by-reference in "stable" will eventually
lose.  Value-add point for cd distributors, I guess :-)