Re: [HACKERS] libpq questions...when threads collide

> > > Either we should keep the current docs
> > > or the release docs online - not both.
> > I disagree, because they serve different audiences. The snapshot docs
> > are very useful to developers, particularly those of us who don't have
> > SGML tools installed but still want to know whether the docs we
> > committed recently look right or not ;-). Meanwhile, current-release
> > documents are clearly the right thing to provide for ordinary users.

Vince, I'm with Tom on this one, having both would be great. The
"developer's only" posting is a holdover from the first days when we
could generate docs on the Postgres machine, and I only had one place
on the web page I could put docs. But having the release docs posted
from the "Documentation" page and the current tree docs posted either
there or on the "Developers" page would be great. I'm happy to
redirect my nightly cron job to put the output somewhere other than
where they are now.

> Um, you mean you commit docs before you know whether they even "compile"?
> As I see it, if you want to edit the docs, you should test them with your
> own SGML tools. With recent sgmltools packages, this is not so hard. At
> least the patch applicator hopefully does this.

No, testing doc output has never been a prerequisite for submitting
and committing doc improvements/updates. If the submitted sgml code is
a bit wrong, the nightly cron job halts in the middle and the output
tar files and web page copies don't get updated. I see the results in
the cron output I have sent to my home machine, and usually fix the
problem within a day or two (would be longer recently since I'm so
busy, but the scheme still is working...).

The important thing is getting the words updated in the docs, and
running jade or the SGML-tools wrappers is still too much of a barrier
if it were a prerequisite.

- Thomas

--
Thomas Lockhart lockhart(at)alumni(dot)caltech(dot)edu
South Pasadena, California