<img> could have multiple attributes that link: 
src, longdesc, and usemap.
[download]

Now that makes sense. I bet I could have figured it out if I'd thought a bit longer. I'm so lazy.

Thanks for your help, podmaster, but surely the guy is intending it to be used as a hash?

One useful attribute of it being a hash would be to clobber incorrect HTML where a link had two HREFs.
--

($_='jjjuuusssttt annootthheer
     pppeeerrrlll haaaccckkeer')=~y/a-z//s;print;
[download]

"but surely the guy is intending it to be used as a hash?"

I'm not a mindreader. HTML::LinkExtor is a pretty mature module, and I doubt the interface should/will change.

You can certainly try to persuade the guy (perlsonally i'd rather just write HTML::LinkExtractor which would do all the things you say here, but would also extract the link text (stuff in between <a ..> </a> tags).

" One useful attribute of it being a hash would be to clobber incorrect HTML where a link had two HREFs. "

You don't have to worry about that (when in doubt, test).

use HTML::LinkExtor;
my $p = new HTML::LinkExtor(
    sub {
        print "@_\n"
    },
);
$p->parse( q{
<a href="BUTTER"    href="SCOTCH">
<img src="AND" src="PEANUTS">
});
__END__
a href SCOTCH
img src PEANUTS
[download]

____________________________________________________
** The Third rule of perl club is a statement of fact: pod is sexy.

I don't want to persuade the guy to change. I've long ago had my question answered, i.e. "why isn't there just a single attribute for each tag?", because it's been noted that there are indeed instances where there are two or more attributes which constitute a link in one tag.
--


($_='jjjuuusssttt annootthheer
     pppeeerrrlll haaaccckkeer')=~y/a-z//s;print;
[download]