dakretz said:
>   How accurate is this assessment?

it's half-assed accurate.

mostly because it's looking at the p.g. corpus
from the standpoint of its two major file-types.

but we'll want to look at it from the perspective
of the users who will access it, and how...

(the answer to that is mobile, mobile, and mobil.)

moreover, some of your points, even as given, are wrong...


>   1.  original plain-text.  significant but declining strongly.

no.  significant and _increasing._  (a rising tide lifts all boats.)


>   2.  original .html files.  significant but declining slightly.

no.  significant and _increasing._  (same tide, different boat.)


>   3.  plain-text derivatives.  declining.

dead wrong.  this is the segment that is increasing fastest.

most of the places that make derivatives use the plain-text.


>   4.  .html derivates.  significant and increasingly so.

well, not quite "dead-wrong", but still wrong nonetheless.

the .html files have far too little consistency to be used in
a systematic creation of derivatives, not without glitches...

some places use the .html file, but then "fall back" to the
plain-text version if they see problems with the derivative.

but most can't spend that much energy on quality-control,
so they've resigned themselves to using the plain-text files.

which is not that big of a sacrifice, to be perfectly honest...

indeed, the system giving the most consistently best results
is the iphone viewer-app "eucalyptus", which utilizes _only_
the plain-text files; his converter is giving very good output.

and, to help get people's heads on, and completely straight,
it's good to do the reminder that many of the .html files are
the result of a straight-out conversion of the plain-text file.

and these files, because they're machine-generated, _are_
consistent enough to be used in a systematic conversion...
it's the "hand-crafted" ones that cause all of the problems,
which is something that i first pointed out many years ago.

when problems with the auto-generated .html files do occur,
it's usually due to an underlying glitch in the plain-text file.

so auto-conversion of plain-text is the best way to proceed.

and i've maintained for 7 years now that such a conversion is
not just _possible_, but our best course of action to follow...

for several years after i started, i left my argument unproven,
just to see who would jump at the bait and try to dispute it...

after destroying all that opposition, i have since proven that
it is indeed possible to use a plain-text file as your "master".

why y'all continue to ignore this proof, i simply do not know.

but i'll keep making the case, until all of you can see it clearly.

-bowerbird