OPML pot luck

My mind, I find, works more like a crock pot than a microwave (OK, so now I’ve invited the “crock” jokes — have at it in the comments). I don’t think I’d make a good “scoop” reporter, because my first impressions usually come out half baked. They’re much more palatable if I let the juices stew for a day or three before opening the lid.

Now that the buzz over Share Your OPML has quieted down a bit, I’ll serve up a few of my braised thoughts on it.

I like the concept. It’s nice to know who is subscribing to your feed, or a feed you like to read. It’s handy to get suggested reading from people you respect. It could be informative to see who else reads the same sorts of things you do. It’s interesting to see which feeds get the most readership. If only all of this really worked.

By definition, the only results you can get from this service come from subscribers who have uploaded their OPML and shared it. That heavily skews the sampling in favor of the technically savvy and blog-obsessed. Perhaps that group largely overlaps feed subscribers in general, but it hardly includes all blog readers. In my own case, I suspect that my post on the limitations of OPML just four days before Share Your OPML went public may have kept some subscribers from uploading their subscription lists, even though I stated in that post that OPML does serve the purpose of exchanging subscription lists quite well. Additionally, many blog readers do not subscribe to feeds. Believe it or not, not even all techies read blogs daily.

The solution to that limitation would be to have some other, more automatic means of uploading lists. For instance, if WordPress and other blogging platforms included the option to upload your BlogRoll automatically, then we’d get lists from people who don’t even know what OPML and RSS are. Of course, the upload would then have to discover the feeds from the sites listed.

But given the current limited sampling, I still have problems making use of the information, mostly due to URL literalism. Share Your OPML differentiates feeds by URL, including all parameters. That means that if you have a feed that can be accessed by two different URLs, your numbers will be split between them. My feed has two URLs, because I migrated it to WordPress under a new URL, but made the original URL spit out the same content for compatibility. So half of my subscribers are listed under each URL. Many other sites provide RSS 2.0, RSS 1.0, and/or ATOM feeds — each with their own URL — so their stats are split also. When TDavid looked at the stats for his RSS 2.0 feed, he must have wondered “where’s Sterling?” until he found me under the ATOM feed stats — somewhere along my migration of feed readers that one got discovered by default.

So how could Share Your OPML fix that? One way would be to have feed owners claim their feed, and then enter synonym URLs for disambiguation. That would probably work best. Or, Share Your OPML could limit the key to the base URL of the feed, but that wouldn’t help cases where two URLs serve the same content (mine, for instance). Or, they could compare the content of the feeds at any given moment, but that’s probably too server-intensive. Any other ideas, folks?

Even though I’d like to see the site stats combined, I would also like to still see them subdivided by URL. That lets you know how each subscriber got to you.

Next problem: Share Your OPML is severely search-challenged. I would like to be able to search for a feed by name, rather than URL. Seems like a little work with Google site search should help with that. I’d also like to be able to search for a subscriber by name, rather than having to find that person in the subscriber list for a feed that I think they should be reading (like their own).

If a subscriber has a feed of their own, I’d like to be able to see that feed’s URL so I can check out their blog and, who knows, maybe subscribe to it in return.

Hmm posted on Monday about how the site was being “gamed” by artificially long subscription lists in order to get a mention as “top subscriber”. As I said in the comments, another possible exploit would be to create multiple accounts in order to inflate the subscriber numbers for a particular site. All that’s required is a unique e-mail address, which a bot could easily manufacture. When I signed on, the verification e-mail required no response. I would suggest adding that requirement, as well as perhaps CAPTCHA or some other manual step that would make such an exploit difficult for a script, and temporally prohibitive for a human.

Anyway, I like the idea. It just needs a little more cooking time.