Opened 11 years ago

Closed 10 years ago

#284 closed editorial (wontfix)

merge p1 (messaging) and p2 (semantics)?

Reported by: fielding@… Owned by: julian.reschke@…
Priority: normal Milestone: unassigned
Component: non-specific Severity: Active WG Document
Keywords: Cc:


It is easier to read and make use of a specification that is presented as a story instead of as a long list of boring reference information, since implementers want to focus on the task at hand rather than the entire protocol.

Ideally, HTTP should be defined progressively as we describe how to initiate a request, how to understand a received request, how to forward a request, how to respond to a request, how to send a response, how to understand a received response, and how to forward a response. That allows each type of implementation (user agent, intermediary, origin server) to be explored and the protocol defined without reference to all of the other implementations.

Unfortunately, I am finding it hard to do that with the split between messaging and semantics. The user agent story needs to be understood in order to explain what the target resource is and thereby define what ends up being the request-target, Host, and where the connection is made. But the story is all about semantics.

Likewise, p2 is currently just a list of protocol elements and field definitions. There is nothing tying it together with *why* and *when* these elements are used because that motivation is almost entirely governed by URIs and connection management (in p1).

My inclination is to try merging p1 and p2 for -14 and perhaps move all of the IANA templates into a new p2. Comments?

Change History (7)

comment:1 Changed 11 years ago by mnot@…

I'm wary of such big structural changes, especially as it's still not clear that the relationship with the other sections won't need similar tweaking.

Also, I've always seen having the current p1 as a separate doc good, in that we can specify a new wire serialisation / binding (how I hate that word) easily, without affecting semantics, etc.

However, I won't lie down in the road about it.

comment:2 Changed 11 years ago by mnot@…

  • Owner set to fielding@…

Prague discussion: try moving "Request" and "Response" sections from p1 to p2

comment:3 Changed 11 years ago by julian.reschke@…

  • Milestone changed from 14 to 15

comment:4 Changed 11 years ago by julian.reschke@…

  • Milestone changed from 15 to unassigned

comment:5 Changed 10 years ago by fielding@…

From [1583]:

Split connections into forwarding of messages by intermediaries and connection management (relocations and minor wording additions only). Move definition of Connection, Via, and Upgrade into connection management. Related to #284.

Remove miscellaneous notes and now-empty header field section.

comment:6 Changed 10 years ago by mnot@…

  • Owner changed from fielding@… to julian.reschke@…

Paris: Julian to re-shuffle bits in p2 to reduce # of sections; revisit afterwards.

comment:7 Changed 10 years ago by mnot@…

  • Resolution set to wontfix
  • Status changed from new to closed

P2 seems to be becoming the "HTTP Core Semantics" section, whereas p1 is the wire serialisation (which needs to somewhat separate if we want to do HTTP2. Closing.

Note: See TracTickets for help on using tickets.