Bloglines Web Services trades one bottleneck for another
Last night Mark Fletcher spilled the beans, and today we have a press release: “New Bloglines Web Services Selected by FeedDemon, NetNewsWire and Blogbot to Eliminate RSS Bandwidth Bottleneck.” Let’s take a look.
Does this solve anything?
I’ve been a huge fan of Bloglines for a couple of years now, but this “solution” concerns me, because the RSS bandwidth bottleneck is now Bloglines. Once more, polling does not scale, and we haven’t solved that problem at all. What Mark is saying is the following: “Content providers are currently complaining about the load imposed by N RSS aggregators polling their servers. So we’ve set up a service that allows those aggregators to poll us instead. We’ll poll the content providers, in turn, but responsibly and as infrequently as possible.”
What this means, in practice, is that content providers will indeed see fewer polls as more end-user apps start polling Bloglines to get that same data. But now bloglines.com will have to satisfy N × C amount of load per unit time: the number of aggregators times the number of feeds. Of course, this is really on the order of N × C; there are some hidden linear factors there, since not all clients read all subscriptions. And it’s possible (likely, since Mark is a smart guy) that Bloglines will send customized sub-feeds to each client, containing only the information that that client is missing, so this will reduce the overall bandwidth associated with this load.
But what we’ve done here is trade one kind of bottleneck for another. It’s not just the bandwidth issue that concerns me, either: this solution is predicated on a dangerous dependency on the existence (and freedom) of the Bloglines service. When Bloglines shuts its doors, or holds out the collection tin, your choices will be to:
- pay up, or
- start polling feeds directly again
…and now we’re back to the bandwidth problem. The point I’m trying to make here is that computer science has made progress in dealing with distributed systems like this, and this solution doesn’t seem to take advantage of any of that.
More later, when I’m not trying to finish a paper critique for class.
Background.
If you hadn’t heard about the recent RSS bandwidth flap,
it gained momentum again with a July InfoWorld piece on RSS Growing Pains. Then, earlier this month, MSDN started retracting some RSS feeds, citing bandwidth issues; Scoble elaborated. There’s a lot of other conversation on the topic, too, but most of it assumes that polling is here to stay, and that the way forward is to make the polling as lightweight as possible (with strict If-Modified-Since adherence, etc.).
Coverage elsewhere on the net.
TechDirt: building the RSS platform; Pedro Melo offers cautious hope; Ed Goodwin also sees the business angle; (more later)
Update 16:30: O’ReillyNet weighs
in with an analysis. I continue to appreciate Mark Fletcher’s
fondness for dirt-simple Web APIs; I continue to recoil in instinctive
horror at the way this concentrates traffic at Bloglines.