Ilya Zakharevich <[EMAIL PROTECTED]> writes:
> On Fri, Feb 04, 2000 at 10:14:44AM -0800, Russ Allbery wrote:
>> They're missing features. Neither pod2text nor pod2man have ever done
>> this to my knowledge. (I really dislike this corner of POD semantics,
>> BTW. Interpretation of an L<> directive should not depend on knowledge
>> of all of the headings in the document.)
> Well, I would substitute "should not" by "should", and everything
> becomes OK. ;-)
> You optimize for a writer of pod2xxx. POD is optimized for writers of
> POD.
Precisely, which is why this behavior is broken. If I say something is a
pointer to a section heading, I expect the translator to deal with that,
even if the section heading doesn't (yet) exist. I may be doing this for
a reason. A lint checker should pick up the fact that the section header
doesn't exist, but the meaning of what I write should *not* change as I
add more sections. That's just wrong, IMO. I don't know of any other
markup language that does this.
It doesn't help that the current L<> semantics are fragile in the extreme,
and that it's not clear what the available different types of links are.
What we have currently is a clear-cut ambiguity with no hint at all in
perlpod as to how the ambiguity is resolved:
L<name> manual page
L<"sec"> section in this manual page
(the quotes are optional)
So if I say L<POSIX>, is it a link to the man page about the POSIX module
or is it a link to the section in the current document about POSIX
compliance in whatever I'm writing about? How do I tell the POD
translator unambiguously that it is a link to the POSIX manual page, *not*
to any section in the current document?
--
Russ Allbery ([EMAIL PROTECTED]) <URL:http://www.eyrie.org/~eagle/>
