[CALL FOR REVIEW] doc and www converted to XML

[Simon L. B. Nielsen]

On Wed, Aug 29, 2012 at 3:52 PM, Warren Block <wblock at wonkity.com> wrote:
> On Wed, 29 Aug 2012, Hiroki Sato wrote:
>
>> Gabor Kovesdan <gabor at FreeBSD.org> wrote
>>  in <50392481.1070708 at FreeBSD.org>:
>>
>> ga> Em 25-08-2012 19:58, Hiroki Sato escreveu:
>> ga> > Ah, what I wanted to point out is that not only in the headers the
>> ga> > ids are converted to lower case in the new framework.  I also
>> ga> > think we can ignore the impact for the headers.
>> ga> >
>> ga> > As you know, the Jade program uses the upper case because it
>> ga> > supports SGML general name normalization for the ids and NAMECASE
>> ga> > GENERAL is YES in SGML declaration for HTML.  The xsltproc program
>> ga> > does not always do so because NAMECASE GENERAL is always NO in
>> ga> > XML.
>> ga>
>> ga> Ok, now I see you are referring to docs not www. I was talking about
>> ga> www because I only did manual conversion there. A side effect of
>> ga> changing the SGML declaration used by Jade to enforce XML syntax is
>> ga> that NAMECASE GENERAL is NO as you said. For now, I don't know what to
>> ga> do with that. If we start using XSLT, we can customize the DocBook XSL
>> ga> stlyesheets to do such a normalization. I think it shouldn't be
>> ga> difficult but I'll have to see how the corresponding parts are written
>> ga> in DocBook XSL. Do you have any suggestions how to handle this for the
>> ga> interim period?
>>
>> If we continue to use the upper-case ids, converting all of id attr
>> and xref's linkend attr to upper-case in the source file, not in the
>> style sheet, is needed.  Case normalization of ids by XSLT is wrong
>> because relationship between xref's linkend and id must be correct in
>> the XML source files.
>>
>> I am still wondering if we should stick to the upper-case ids, but
>> considering consistency and compatibility, using upper-case is the
>> safest.  Our possible options in my mind are:
>>
>> 1. Convert ids to upper-case and use them.  No compatibility problem.

What are the arguments against this? It's at least the simplest one...

>> 2. Use lower-case ids.  Links from pages outside of www.FreeBSD.org
>>    may be broken.
>
> Is it practical to generate small static "This link has changed" pages for
> legacy uppercase links?

That does sound a bit painful... many small files around for a long
time. The redirect is better, though it won't work for mirrors....

> Generating a list of "RedirectMatch permanent" lines for manual addition to
> the webserver config seems more elegant.  After the switch was made to all
> lowercase, the list of obsolete uppercase links would not grow.

How many are we talking about? I would prefer not having many
redirects which have to keep around forever.

-- 
Simon L. B. Nielsen