HTTP batching
February 11th, 2008
John Panzer has an interesting post up concerning batched HTTP requests. You can almost taste John’s fear of stop of energy. 
That’s pretty reasonable–I bet there will be a lot of dogmatic resistance to the idea. Objections of the along the lines of “I don’t think you need that” can be safely ignored. But there are other types of comments that are worth considering.
John compares the proposal to HTTP pipelining, but that’s a little different because pipelining only allows “safe” requests such as GET. If you start doing an opaque POST request that updates a bunch of different resources, that’s going to harm client and proxy cache coherence, unless they are wise to this new tunneling mechanism. HTTP as she are spoke has many of those, but this would be yet another. Maybe cache invalidation on POST/PUT/DELETE is already so broken that it doesn’t matter, but I’m concerned. And hey, I’m doing my part in the Battle Against Latency.
The imagined objection to MIME is pretty irrelevant. HTTP already uses multipart MIME messages for range responses.