The business of RSS

How do you count subscribers in the RSS network? Tim Bray meditates on the question in an essay on the subject. Dave Winer says that Radio's Web Bug Simulator (WBS) solved the problem last year. There a few different issues here to tease out, but in the end I'm not sure there is, or ever was, a problem.

As it's currently used, the WBS is more about transparency than comprehensive measurement. Consider this report, which tells us who has fetched Graham Glass' RSS feed from Radio UserLand1 today. As you can see, I am one of his subscribers. The WBS technique makes Graham's subscribership transparently visible to the world -- an interesting situation that was the focus of my comment which Dave cited. Now as it happens, Graham could otherwise know that I subscribe to him, since my channelroll (which Yoz Grahame has cleverly ascended to the top of -- nice hack!) already makes that fact known to the world. But if I didn't choose to publicize the fact, you could still piece together my subscriptions by crawling through the WBS reports. And if you'd rather not publicize your reading habits, you can disable the WBS. The transparency effect is really cool, and is essential to the constitution of blogspace in ways I think none of us yet fully understand.

The WBS technique could be used to measure subscribership for commercial purposes. Other measurements exist too. In RadioSpace, the RSS Hotlist reports the top 100 most-subscribed-to RSS feeds. Note that as of today, my blog appears twice: at #37, and again at #100. See The RSS Hotlist: quantity vs quality for background on this. Briefly, when I moved my feed, I couldn't redirect subscribers to the new address. (When I moved again, I learned that happily a solution now exists, at least for Radio and NetNewsWire.) Back in November, I had 251 Radio UserLand subscribers to the new feed but 145 were still fetching the old one. A few of the 145 have dropped away, but today the numbers stand at 365 and 125. This means that 125 "subscribers" have continued to fetch my old RSS feed, which hasn't changed in over 6 months. Who are these "subscribers"? In this case, it's users of the Radio UserLand aggregator2 who haven't noticed my feed go dark. It's hard to notice the absence of something. If one of your 100+ feeds goes dark, would you notice? If it's one you care about, yes. Otherwise, no.

Robots, of course, don't care about anything, and the vast amount of RSS fetching is robotic. I submit that the WBS technique, if widely implemented, would become largely robotic too. As Steve Gillmor points out, it becomes a question of who's subscribing. And, I'd add, who's responding. There are already well-established ways of figuring this stuff out, and they play beautifully into the existing weblog/RSS architecture.

It'd be useful to see something other than generic aggregator signatures in server logs, so the ideas that Tim was kicking around, and/or a WBS-like approach, would be helpful. But even if such information arrived, nobody at InfoWorld would much care at the moment. Our webstats people basically just ignore all rss.xml page-"views" because robots aren't interesting. What's interesting is people who respond to the feeds. And they respond, in the time-honored fashion, by clicking through to the website, thereby displaying ads. Because our CMS decorates the RSS URLs delivered in the feed (as it decorates URLs in different areas of the website), the RSS-referred views are known as such.

How do we know who's subscribing? We don't, yet. If and when we want to attach value specifically to an RSS feed, though, the mechanisms for doing so are again -- I think -- well-known. Currently, for example, this blog comes in three flavors. You can read it on the web, along with an ad. You can receive RSS blurbs that invite you to click through to the website -- again, displaying the ad. Or, as I mentioned yesterday, you can take the full feed in XML, disintermediating the browser. Few folks do that. If more chose to, it could become a paid feature, or a free benefit to InfoWorld subscribers.

The relationship between UserLand and the New York Times illustrates how this might work. When you install Radio UserLand, you're offered a special set of Times newsfeeds, with more stuff than you'd usually get. These are password-protected and not available to the general public. Inside Radio, there's a table of URLs that look like this:

http://radiouser:xxxxxxx@partners.userland.com/nyt/arts.xml

In other words, Radio encodes the name/password credentials for these special feeds. As I understand it3, this is effectively a group credential shared by all RU users. There's a relationship between UserLand and the Times, not between individual RU users and the Times. But it's easy to see how, using standard e-commerce techniques, the Times could arrange to invidualize its relationships with RSS subscribers.

I believe this kind of thing will happen. So far as I can see, though, there's prior web art for all the pieces of the e-commerce puzzle. There is a ton of innovation that can and should still occur around blogs and RSS. But are innovations needed purely to enable a blog-related business model? Maybe, but if so I'm not sure what they are.


1 I think other aggregators could show up here, but currently don't.

2 Because only RU feeds this data to the hotlist, right?

3 Again, please correct me if I'm wrong.


Former URL: http://weblog.infoworld.com/udell/2003/05/28.html#a705