Re: Working Group Last Call: Compression Dictionary Transport

On Fri, Jun 14, 2024 at 1:37 PM Watson Ladd <watsonbladd@gmail.com> wrote:

> I support publication. I have however found two nits/lacuna in my
> understanding:
>
> 1: It's not clear to me that RFC 7932 actually supports user specified
> dictionaries. It's obvious what that support should be, but it isn't
> specified. Maybe I missed it and we should add a sentance somewhere or
> a reference.
>

Thanks. It's the cited shared-brotli draft that defines how user-defined
dictionaries are used with brotli streams. I've asked the team if they can
see about making it a stable RFC since the draft has expired. The actual
format and encoding sections for the stream format reference the draft
rather than the underlying brotli RFC.


>
> 2: In Section 5 I think we should add for clarity "The dictionary is
> treated as a raw dictionary as per section 5 of RFC 8878".
>
>
Thanks. I added that as well as made it clear that both of the defined
content-encodings are explicitly for the "raw" dictionary type with a link
to the "type" section of the "Use-As-Dictionary" response header where the
dictionary types are specified. Specifically, only "raw" is currently
defined but it does allow for forward compatibility if new dictionary types
prove to be valuable (which would require specific content-encodings that
are linked to the other dictionary types).

Hopefully that isn't too confusing but things get more complicated when you
want to use a dictionary format that is specific to a compression format.
In that case all sides need to know that it can only be used with that
specific format and it will use a different content-encoding to keep things
clean.

I updated the repository with the clarifications:
https://github.com/httpwg/http-extensions/pull/2809/files

They'll be picked up in the next draft release/publication after we get a
chance to gather any other suggestions.