There are a few implementation issues depending on just how clever you want to be specifying a reference. Finding [id://nnnnnn] references would be easy. Finding <a href="xxxx">that node</a> references could be a bit mixed - it may be a bit harder to dig out the actual reference. Resolving [node name] references may be impossible - there are duplicate node names. Resolving "you know, that node by thingy, the wizard fellow, about sorting stuff" may well be beyond most humans let alone some code, but is still a reference, of sorts.

A second thing to think about is, when should nodes be parsed for references? Each time they are updated in the database? That gets a bit tricky avoiding bumping reference counts the wrong number of times. Something the vote fairy should do? She may get a bit grumpy at the unacustomed work (and you better watch out for the Vote Fairy Union too). Should all the old nodes be processed, or only new ones?

If a node is referenced multiple times from a single node does that count once, or once each time?

Having said all that, aside from the implementation issues, I quite like the idea. More nice warm fuzzies handed around and more opportunity for statistical analysis. Fun :)

I'd say the [id:] and <a> tags would be sufficient. As for  [node name] and duplicate names, how does the system handle that now? What node does it pick when I click on such a link? For counting references, I'd say if the reference was unambiguous, count it. Otherwise, no.

And yeah, the "vote fairy" could audit all updated nodes once a day and update any reference counts on other nodes. If the vote fairy is already too busy, hire a reference trawl (troll). :)

Multiple references to the same node in one node would count as one reference.

Yes, process all old nodes. Eventually. :) You could do, say, 5000 a day.

At present a link such as Re: Ook interpreter takes you to the duplicates found page with a list of matching nodes.

---

DWIM is Perl's answer to Gödel

What do you think?

I think people can edit almost all of their nodes at almost any time, so unless you update the reverse index for every node with every change, you can't be sure your information is accurate.

Because knowing what other nodes link to a node is an indicator that those nodes are about similar subjects.

Yes, that and one can also see it as a measure of popularity/quality. Making Super Search able to sort by number of references would be a nice feature I'd use. A bit like google page rank.


holli, /regexed monk/