>> Notice the <i> should have been turned into &lt;i&gt; but wasn't.
>> As a result, a spurious begin-italics code appears and three
>> characters of the PR disappear.

> The problem with this is, "how do you do it correctly?"

Well, define "correctly" to a sufficient level of precision and the
problem will disappear. :-)

> It's nice to allow people to include html, e.g. anchors to their home
> pages' URLs, in PRs, and have them come out as real html on the Web.

Hm.  For human browsing, just displaying on a screen and being looked
at, perhaps you're right; it depends on whether you find it more
important to show the PR contents "correctly" or to show HTML in
people's signatures "correctly".  And for human browsing it does make
some sense to prefer the latter.

But the problem is that this is crippling for anything that wants to
just use the Web engine as a way to retrieve PRs.  Unfortunately the
only other mechanism available, as far as I can tell, is mail to
query-pr.  And that has problems.  The most crippling one, the one that
got this discussion started, is that it won't return closed PRs.
(Secondarily, it dumps the PR into my mailbox rather than producing it
on stdout; tertiarily, lines beginning "From " get mangled, though
admittedly that is the fault of the Berkeley mailbox format rather than
the fault of the query-pr response software.)

It may be as simple as providing a flag one can pass to the Web query
software to disable all the HTML hacks, just slapping <pre>...</pre>
around the PR and doing &xxx; mapping globally; it might be better to
do some other query mechanism.  I'll even volunteer to maintain the
software, given read access to the PR database.  If you have stats on
how much load it takes, so I can get some idea what I'm getting into, I
may even volunteer to run the thing, though that'd mean having to keep
another copy of the PR database in sync with the master one.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu