There has been a lot of discussion lately about RSS bandwidth use. This morning I came across Bob Wyman’s RFC3229 with Feeds. Instead of getting an entire feed, RFC 3229 would extend HTTP to allow sending deltas instead of the whole thing. I’ve got mixed feelings about this idea. I don’t think that standards (like HTTP), should be pressed in stone and never change, but by the same token changes made to it have to be taken very carefully. I haven’t read through everything on the ideas behind RFC 3229, but my gut feeling is that this seems a little on the strong side to solve a bandwidth problem for feeds.
Only time will tell what ideas really catch on to deal with feeds and the load (both system and network) they impose, but I suspect that the ideas most likely to be adopted will be the ones that have the lowest work to benefit ratio (the least amount of work for the most benefit). It may be that there will be two or three likely winning ideas in the end: server side, client side and both. Extending HTTP would fit into the both category and seems the least likely to catch on because both sides would have to implement it in order to see any benefit.
One reply on “Extending HTTP”
Joseph, if it makes you any more comfortable, RFC3229 was issued by the IETF as a “standards track” document in January of 2002 — almost three years ago. I’m not really proposing anything new other than a new instance-manipulation method which is feed specific. In part, my proposal is motivated by a desire to focus people on the standards have already been worked out at great expense in the past. We should do the best with what we have before we move on to defining completely new stuff…
Personally, I’m confident that the “real” solution to this problem is the kind of publish/subscribe “push-based” system that we provide at PubSub.com. However, before trying to move people off the technology they use today, we think it makes sense to make it as good as it can be. Once we’ve seen the true limits of pull, we’ll be able to appreciate push better and have a better understanding of how to implement and use it. If we implement RFC3229, in combination with all the other existing HTTP methods like conditional-get, compression, etc. then we will have taken polling or “pull” just about as far as it could possibly be taken…
bob wyman