- From: James M Snell <jasnell@gmail.com>
- Date: Tue, 15 Jul 2014 10:41:57 -0700
- To: "ietf-http-wg@w3.org" <ietf-http-wg@w3.org>
I really haven't had the opportunity to go through an read all of the messages to understand the full context or what's actually being written into the spec so I want to make sure... I'm seeing quite a few comments passing by about pushing things into http/2 "extensions". While having a clear, well-defined extensibility model in HTTP/2 is a good thing. extensions themselves are a certainly a double-edged sword if not handled appropriately. There are generally two points of view: 1. Extensions are good because they allow the protocol to evolve and grow to meet new demands over time. and 2. Extensions are bad because they give vendors excuses to not implement functionality they disagree with. I.e. "We don't actually want to do it that way, so we're going to pretend to be 'open' and 'encouraging' by saying that can be an "extension" when what we're really going to do is kill it off by not implementing it". So the question is: in HTTP/2, are there going to be clear notions of mandatory-to-implement-but-not-mandatory-to-use extensions or are all extensions going to be optional-to-implement. The difference is critical. For example: I typically do not spend my time building HTTP/2 infrastructure; I do, however, build applications on top of HTTP/2 infrastructure. And if one of my applications depends on the ability to use a particular HTTP/2 extension, I want to be clear on whether or not I can rely on that extension being implemented or not. - James
Received on Tuesday, 15 July 2014 17:42:44 UTC