[dispatch] Observations on process after reviewing the iotl draft

[Robert Sparks <[e-mail]>]

All -

I just sent in a gen-art LC review on the iotl draft. While putting that 
review together, I went back through the email discussion of the draft, 
and when taking it in all at once like that I saw a few things I think 
we should discuss.

First, Christer has worked very hard to do the right thing, and to get 
the IETF community to do the right thing with this draft. Thank you 
Christer. Please keep doing that.

Our response as a community was underwhelming, and very slow. I don't 
think we need to beat any individuals up about that, but I do think we 
have an opportunity to study how this went, so we can do better with 
other things.

There were concerns raised early (last May) about the design (URI 
parameter vs header field parameter) that I don't think were well 
resolved on list. When that discussion came up, I think we should have 
taken it as a flag that this would have benefited from a WG discussing 
it with a chair driving the discussion. When we decide to send something 
off to AD sponsorship in the future, the AD should watch for similar 
conversations and send it back to us (and we should help the AD of the 
time recognize what's happening). Again, I'm not trying to call what's 
in this document into question at this point. I just wish we had done 
better back in May.

Much of the discussion danced around URI parameters requiring 
standards-track, and header parameters only requiring informational, 
rather than which of those things (or something else) might have the 
most technical merit (and based on the conversation, there were 
implementors vested in the choice that had been made before the draft 
came here). As we've often seen, this points to bringing the problem 
here earlier, rather than the solution.

It also points out that the current standards-track requirement for URI 
parameters is probably overkill. I'd like to suggest we change that to 
specification required with expert review. The expert would not allow 
parameters that changed the semantics of the protocol unless the 
specification happened to be an IETF stream standards track document.

It was an interesting experiment using the dispatch list to refine the 
document's technical content. In my opinion, we should not do that again.

RjS