Re: bad filenames when creating html/docbook using pretest

[Patrice Dumas]

On Thu, Nov 22, 2012 at 05:13:34PM -0800, Per Bothner wrote:
> 
> I would suggest letting '-' be unescaped.  Yes, there is a small
> chance of a clash.  But you already have that risk, at least on
> case-folding file systems.

Not really, the anchors are always unique.  What could be done would
be to keep - as is in file names.  But having unique anchor name is a
feature of the specification that should not be dropped, in my opinion.

> For myself, I run a sed script after using texi2any to generate
> docbook, before generating html using the docbook style sheets.
> So adding an extra rule -e 's|_002d|-|g' is a simple if not
> 100% robust solution.  

There is a hook for the node file names, and a hook to 
modify node targets and id.  Function references are called 
node_target_name and node_file_name.  But the API is still
experimental.

> More critically, anyone using texinfo for non-English text may
> want non-Ascii letters in node names.  Perhaps allow any
> "NameStartChar" (as in the XML specifications) without escaping.
> This makes for much nicer file names and URLs to people using
> these languages.  True, %-escaping comes into effect, but this
> is handled automatically by browsers and servers.  Let's defer
> this to the standard mechanism.

File names, in the default case, have letters transliterated.  Are 
you saying that there may be some non ascii characters in docbook ID?  
Is it clearly stated somewhere?

-- 
Pat