The Problem with nofollow
As the blogosphere toasts itself over the collaboration that went into nofollow, I can’t help thinking that it’s way too soon to declare victory.
Dare points out that it was a bug for vendors to assume that all links are links of positive mention. But the point is that an href never provided any additional metadata about the disposition of the person doing the linking. The only metadata that could be inferred from a hyperlink was ?the person authoring page X found page Y to be sufficiently interesting to link to it?.
Now with nofollow, you have one tiny extra bit of metadata. ?The person authoring page X wants you to be able to vist page Y, but can’t decide if it is interesting or not?.
I think it would have been much better (and just as easy) to add a ?rating? attribute instead. Rating would capture the opinion of the page author about the site being linked, and might be one of ?Cool|Unknown|Spam|Crap?. Search engines could ignore links with unknown rating, and actually downrank links reported as spam.
The problem with nofollow is that it’s only effective as a deterrent if 90% of sites with comments implement it. It costs the same amount of money to spam a nofollow site as a non-nofollow site, and it doesn’t cost much more to spam 10,000 sites than to spam 100. If 5% of the sites you spam implement nofollow, you just spam 5% more sites for the same ROI at a tiny fraction of the incremental cost. I don’t see this acting as a deterrent.
On the other hand, if the search engines actually allowed downranking, then spammers would very quickly be punished. Again, it wouldn’t work to downrank all items that appear with ?nofollow?, because most of them will be legitimate (especially if your system works). So you don’t want to punish legitimate links.
In practical effect, nofollow seems like a really crappy hack that tells search engines ?this link probably comes from a comments page?. For starters, the search engine should be smart enough to figure that out. And furthermore, it doesn’t help much at all. And finally, I should be able to search on this metadata just like any other (for example, I can search for pages in a date range that is likely to get filtered out by relevance rank; why not be able to search for all pages with a rating of spam?) It seems like it’s actually more work for search vendors to take this little page rank hack than it would be to treat it like any other metadata.
If nofollow is the thing that stops comment spam cold (or slows it down noticably) then I’ll be pleased. But I’m very skeptical right now.