Re: PATCH: produce xhtml output and get rid of evilsedhack

[James Henstridge <james jamesh id au>]

Curtis Hovey wrote:

> > Some downsides include:
> >
> >   1. if no <meta http-equiv="content-type"> tag is in the source HTML,
> >      libxml will assume it is latin-1.
> >    
> The content should be UTF-8 (or at least it was when I made some mass
> updated last year).  Do I need to add the meta tag?  I believe point 2
> from above (and the fact that latin1 is a UTF-8 subset) that nothing
> needs to be done, save for recent content that might be localized
> improperly.
>  
Latin-1 is not a UTF-8 subset.  Anything above codepoint 127 is 
represented differently, which is the problem.  So if a page without the 
meta tag contains 8-bit characters and it is in UTF-8, it would get 
mangled (eg. "®" would come out as "®").

Perhaps the right solution is to just add the meta tag to every .wml 
file ...

> >   2. makes the website build depend on libxml2/libxslt
> >    
>
> I don't believe this to be a problem..  Will order realse from the past
> year or two suffice?
>  
I don't think I'm using anything new or uncommon.  The exslt date 
functions have been in libxslt for a long time, and everything else is 
standard XSLT.

> >   3. is a little more resource intensive than sed (not enormously though).
> >
> > (1) was an obvious problem for the translated versions of the Gnome 2.10 
> > press release, which I've corrected in the attached patch.  I haven't 
> > checked to see if other files are affected.
> >
> > While this isn't adding much new right now, it should also make it 
> > easier to customise things in the future (it should also result in less 
> > of those "optimised for standards" links spitting out lots of errors ...)
> >    
> We use evilsedhack in foundation too.  Do you anticipate portability
> problems if we upgrade all evilsedhacks?
>  
Nope.  It would be worth going over any of the warnings printed by 
xsltproc though -- these usually indicate invalid HTML, which can result 
in unexpected results if it is particularly bad.

James.