"Manuel Collado" <m.collado@domain.invalid> wrote in message 
news:joqqmr$ki4$1@speranza.aioe.org...
> El 11/05/2012 17:32, Simon Wright escribi�:
>> Jacob Sparre Andersen<sparre@nbi.dk>  writes:
>>>...
>>> If I could write<http://www.adaic.org/rm05/9.7.3#6>  and people would
>>> end up at paragraph 6 in section 9.7.3 of the Ada05 reference manual, it
>>> would be much easier than linking to the whole section and then writing
>>> that people have to scroll down to paragraph 6.
>>
>> I'd have thought that would be straightforward enough .. the paragraph
>> numbers are in a div with class paranum. A simple patch means I can
>> write<RM-A-3-2.html#p13/2>  for para 13/2 of the 2005 manual (I'm not at
>> all sure of the legality of that slash, but Safari and Chrome are 
>> happy!!).
>
> Is seems that p13/2 as a "name" target is legal in HTML, but not in XHTML. 
> In addition, IIUC the use of the "name" attribute is deprecated in favor 
> of the "id" one, and XHTML5 even has dropped it.
>
> "id" values don't allow slashes. But slashes could be easily converted to 
> hyphens or underscores, if browser standard compliance really matters. So 
> we could write <RM-A-3-2.html#p13-2>, for instance

We don't need the slashes at all, since that just a version indicator, and 
the root path specifies the version of the document anyway. I was just going 
to strip them when I added them to ARM_Form. That's especially valuable 
since these versions are much more likely to change than the paraagraph 
numbers themselves, so it would make the links more resilient to change 
(important for working versions of the Standard).

I was more worried about the periods (which are not optional). They could be 
converted to something else, but then the correspondence between the anchor 
name and the paragraph number would not be as obvious, which would possibly 
hurt the usage (when people want to hand-create these links - which seems to 
be the main benefit of the idea).

> In addition, if old and new browsers have to be supported, I believe it is 
> also possible to have both "id" and "name" target attributes with exactly 
> the same, duplicated value.

That would seem to be necessary (I've only used "name" because older 
browsers don't support "id" - I rather doubt any browsers are going to be 
dropping support for it anytime soon because it is widely used).

                                                 Randy.