Skip to main content

Early Review of draft-ietf-wish-whep-00
review-ietf-wish-whep-00-httpdir-early-reschke-2024-01-30-00

Request Review of draft-ietf-wish-whep
Requested revision No specific revision (document currently at 01)
Type Early Review
Team HTTP Directorate (httpdir)
Deadline 2024-02-28
Requested 2024-01-20
Requested by Mark Nottingham
Authors Sergio Garcia Murillo , Cheng Chen
I-D last updated 2024-01-30
Completed reviews Httpdir Early review of -00 by Julian Reschke (diff)
Assignment Reviewer Julian Reschke
State Completed
Request Early review on draft-ietf-wish-whep by HTTP Directorate Assigned
Reviewed revision 00 (document currently at 01)
Result Not ready
Completed 2024-01-30
review-ietf-wish-whep-00-httpdir-early-reschke-2024-01-30-00
I have done a quick review focusing only on HTTP related parts, and on general
RFC style.

Style nits:

- please separate artwork for request / responses and tag them with
"httpmessage" - one use of "COULD" (that's not a BCP 14 keyword) - please use
"Section x of [REF]" instead of "[REF] section x" - most artwork is too wide
for plain text rendering - there's no reason to specify an abbreivated title
when the title of the document is already short enough - 6.2.1 has prose in
artwork instead of plain text, thus the warning about RFC 8141 being unused

HTTP remarks:

- "The WHEP Endpoints MUST return an "405 Method Not Allowed" response for any
HTTP GET, HEAD or PUT requests on the endpoint URL in order to reserve its
usage for future versions of this protocol specification." - I think this is
highly problematic. If you want to reserve specific uses for GET, defining
media types for that purpose and doing content negotiation makes more sense

- there is a trend towards being incorrect or too specific with status codes;
for instance, 409 sounds incorrect if the it's really about the endpoint not
being ready; in which case a 5xx makes more sense. Furthermore, requiring 204
is too specific; there's really no reason why an empty 200 reponse shouldn't
work as well. It's also not clear why "307" "MUST" be supported, and 308 is not
mentioned at all.

- don't be too specific with additional requirements on requests/responses, in
particular when they are already defined by HTTP (see, for instance, handling
of ETags and conditional header fields)

In general, I recommend reading https://www.rfc-editor.org/rfc/rfc9205.html and
to revise the document where it either is to specific, or tries to redefine
that HTTP already specifies.