Keith Devens .com

Well, like I <a href="http://www.decafbad.com/news_archives/000142.shtml">mentioned over here</a>, Disenchanted publishes their links back manually. They're automatically discovered, but they check them out in person. So, the links back won't appear until someone over there actually notices it and approves it for publishing. This lets them knock out duplicates and comment on things. My linkback scripts just blindly show referral links

It is automatic, but it waits for two unique visitors to come in. It solved a problem I had with phony referrers, search engine hits, and Web Email clicks.

When I arrived this morning and took a look a the page there was a link to your site there, with 4 inbound visits already recorded. I'm guessing it probably appeared some time this morning.

Phony referrers can be fun. Last year Dave Winer linked to an article of mine, and one of his anti-fans figured out that if they visited "http://www.scripting.com/?dave_winer_is_a_fricking_idiot" and followed the link, then it'd appear on my page.

Ah hah. That answers a few questions for me too. Completely manual referrers would seem like a PITA to me

I understand some people are writing systems that will fetch the linking page automatically to verify the link and perhaps also nab the page's title. I think this is a good idea.

By far the largest maintenance problem I have is the fact that URLs are rarely unique. http://www.foo.com is the same as http://www.foo.com/ is the same as http://foo.com/index.html is the same as http://www.foo.com/index.html?session_id=2893472. Then there are sites that have query-driven display systems, so a link will appear on a "week at a glance" display as well as the page for a specific posting.

A mature system would have a mix of atomated URL canonicalization and aliasing. I've got the aliasing part down already at two levels (URL and link), but all my programming energy is currently directed at a re-build of the site in AxKit, so other improvements won't come until that part's done.

Incidentally, as part of switching to AxKit -- an XML pipelining system -- I'm contemplating other Xanaduish features. "Transclusion" is possible with XPath, for example. That's where a document can be quoted without cutting-and-pasting the text itself; you insert an address for a span of text instead, and that text is retrieved from the original document as the quoting page is viewed. That way, changes to the original affect all derivatives, and a pathway back to the source of the material can be discovered from anywhere down a chain of transclusions. (Imagine using that in a flamewar).

XPath might also make it possible to do an even finer granularization of backlinks. My site can stick a link at the end of invidivual paragraphs that will take a reader back to a referrer (actually, open up a hovering DIV with the list in it), but with XPath you could identify a word or a sentence.

Xanadu -- the longest ever running vaporware project in computer science history -- may be just around the corner. Again :-)

Ooer, funky autoformatting. Those overstrikes are supposed to be emdashes.

Fixed the formatting. If the "" is next to a word character, it's supposed to be a strikeout. If you put a space around it it's fine. Although I thought my formatter makes it have a space before the "" for it to be a strikeout. I'll have to check my regex.

I like the suggestion to require two travels as a way to exclude things like email apps. I'll have to wrap that into my CF referrer system.

Transclusion can also be simulated with inline frames. Check out the inline preview at http://surfmind.com/?q=transclusion (on the right side of search results, the "inline" links)

That implementation cheats a bit but I've got a DOM only version in the works.