Re: [hybi] [Technical Errata Reported] RFC7692 (4982)

[Takeshi Yoshino <[e-mail]>]

On Wed, Mar 29, 2017 at 5:20 AM, Moonchild <[e-mail]> wrote:

> I think you have to be careful here that you're not going to imply that
> a client must support the parameter in its implementation, but merely
>

I tried to avoid introducing a new term "treat" and added "a hint" to the
first paragraph instead to clarify this point. The term hint is well
explained in https://tools.ietf.org/html/rfc7692#section-5 and readers
should be able know that they can safely ignore it as far as all the
requirements are met. But yes, you're right. I lost the original point by
you again here that the statement for client requirements should be
self-contained.


> that the parameter as part of a response must be *accepted* (but may be
> ignored if no implementation is being provided to back this parameter)
>

In this document, the term "support" is really used to indicate that an
endpoint can accept an offer/response with the extension parameter.
Requirements for endpoints that accepted it are described separately.

The main pain point was that if the spec is not clear that a client should
be prepared (must be able to "accept") for an unsolicited
server_no_context_takeover, it leads the interop issue. "MUST support"
would work enough to avoid this badness.

When readers try to understand what's the implementation to provide,
they'll look for that and find one in
https://tools.ietf.org/html/rfc7692#section-7.2.2 section. As far as the
spec is saying that readers need to always pay attention to
server_no_context_takeover, the confusion can be clarified here.

   If the "agreed parameters" contain the "server_no_context_takeover"
   extension parameter, the client MAY decompress each new message with
   an empty LZ77 sliding window.  Otherwise, the client MUST decompress
   each new message using the LZ77 sliding window used to process the
   last compressed message.

The MAY sentence is already implying that point. The Otherwise sentence
corresponds to the "ignoring the parameter" case. What do you think. Can't
this serve enough for clarification?

Thanks