Something that came out during the Vancouver discussions was, that the users
should be counted, not the machines (as far as I understood it).
Therefore I've created http://www.buildd.net/cgi/archvote.phtml (mainly for
the purpose to count the numbers for m68k, where the problem is similar to
mips(el)), but that page can be used for other ports as well.
I already submitted the first list of m68k users to Andreas Barth, who
included the number on http://release.debian.org/etch_arch_qualify.html, so
I guess that page is good enough for the Release Team. Feel free to use it.