Talk about exceptions to storing and updating headers - #337
Conversation
... and encourage their removal.
royfielding
left a comment
royfielding
left a comment
There was a problem hiding this comment.
There's quite a bit of "headers" usage here -- was it written before the editorial changes?
I'd rather it be consistent regarding "header fields" or just "fields" before applying.
|
Updated to use 'fields' as appropriate. This test relies upon the understanding that If that's the case, I think we need to make the appropriate changes in Messaging; @royfielding should we do that here, or in a separate issue first? |
royfielding
left a comment
royfielding
left a comment
There was a problem hiding this comment.
The first MUST is right, but I believe the last three cases apply to both header and trailer fields. IOW, the names in Connection apply to trailers being forwarded as well.
|
The reason
The reason being that 2616 assumes the trailers would be merged as header fields (or simply dropped) prior to forwarding the message. |
|
@royfielding that makes sense, although I suspect even fewer implementations are going to be doing that; might be worth reinforcing in Semantics? In any case, please review. |
| <li>Header fields that are specific to the connection they occur | ||
| &MUST-NOT; be stored. Effectively, this is limited to Proxy-Authenticate | ||
| <xref target="field.proxy-authenticate"/> and Proxy-Authorization <xref | ||
| target="field.proxy-authorization">, since new fields are | ||
| discouraged from being tied to a specific connection.</li> |
There was a problem hiding this comment.
Note that this is dependent upon the resolution of #331, which may not go this way.
reschke
left a comment
reschke
left a comment
There was a problem hiding this comment.
(see comments; also need change log entry; I can update the PR...)
|
OK, I think this is good to go with that last change. @royfielding could you please review? |
royfielding
left a comment
royfielding
left a comment
There was a problem hiding this comment.
Wait, this is not just talking about caches storing fields now. The changes to messaging are reverting a longstanding consensus (and demonstrated practice) that there are no "hop-by-hop" headers other than Connection and those listed in Connection. The listed fields can't be simply dropped - they must be processed according to their semantics, which drops them when applicable.
|
@royfielding see update |
3 participants
Fixes #165