[boulder.pm] problem with HTML::LinkExtor and <applet>

[Walter Pienciak]

On Thu, 14 Sep 2000, rise wrote:

> > Hi, (Jonathan|Jon|John|rise),
> 
> s/\|John//i or (after 'use English;') all of the above with John being the
> only depracated tag (though still legal).
>  
> > The format is that of a relative URI.
> 
> Agreed.
>  
> [URI spec on subject of handling]
> [HTML spec: $codebase->isa('URI') returns true]
> 
> The combination of those two pretty much cinches it.
> 
> > So all that spec-quoting means, to me, that there's no reason for
> > HTML::Parser to be mishandling <applet> URIs:  attribute specs nail
> > them down unambiguously.
> 
> Agreed. I wonder if someone involved in HTML::Parser failed to check the
> same assumption I made.
>  
> > I like your suggestion about doing two passes (the second pass would need to
> > identify the <applet> stuff, re-create the bogus URLs, delete them from the
> > data structures where they'd been added, and then generate and insert the
> > *correct* URLs.  I may go that route.  I'm just lazy, that's all . . .
> 
> It's certainly not pretty. 
> 
> > I may wind up heaving this onto CLPM and seeing what happens.  I can't
> > believe I'm the first person to be dealing with this.
> 
> That might be the best move. After reviewing the spec it looks to me like
> code and codebase are _two_ URIs, the latter of which happens to be used
> to override the default base URI for the former. HTML::Parser is correctly
> parsing this as a pair of URIs but HTML::LinkExtor is not then using the
> codebase to rewrite the code URI with the base. This is most likely
> because HTML::LinkExtor is registering callbacks for code and codebase
> separately. If that's the case then it sees the two as two completely
> different links (which is the behaviour you get).
> 
> I'm going to poke into the HTML::LinkExtor source to see if I can find a
> callback. I'm starting to wonder if this is the expected behaviour (though
> the far more annoying one from your point of view) since you lose a little
> bit of information when you combine the two URIs.

Just a followup (and context NOT snipped, for the benefit of those who
joined the list in the last week or so):

I did heave the problem onto CLPM, and I got a response from Gisle basically
saying yup and that a patch would be welcome.

I sent him a patch 21 Sept.  I don't really want to release it generally,
in case he has some changes, but would forward it on an "absolutely no
warranty expressed or implied -- it could be a BOMB! and not a patch"
basis to individuals via e-mail.

Walter