Comment on draft-ietf-httpbis-sfbis - lower case keys

[Mitar <[e-mail]>]

Hi!

I know I am late in the draft-ietf-httpbis-sfbis process, but after
using for some time now both RFC 8941 and sfbis extensions, I wanted
to share my experience. I really like it, It is great to have a
standard way to pass values in headers to clients. It is great that it
is compatible with how most other standards are doing it. It is great
that it is typed.

But ... the requirement that keys should be only lower case (section
4.1.1.3) really makes it hard to be interoperable with existing ways
to encode data structures. It makes it tricky to write idiomatic data
structures in many languages which use camelCase and then convert that
structure into structured field values. For example, if you want
idiomatic JavaScript objects on both server side (where you are
sending the header) and client side (where you are reading it), you
have to encode it as snake_case or something. Or ask all users
creating the data structure to use snake_case to begin with. But in my
app I want to pass some metadata to the client and I want the app
itself to decide which fields go into JSON response and which fields
into HTTP header. There is unnecessary friction here.

I would really urge that this is relaxed in one of future versions and
upper case ASCII to be allowed as well. I know that this is primarily
cosmetic, but it would make different languages and libraries easier
to work together.

I understand that there are probably strong reasons for this decision
and that this has probably been discussed already, but I just wanted
to share my experiences with in my view unnecessary friction.

Thank you for great work on the sfbis,


Mitar

-- 
https://mitar.tnode.com/
https://twitter.com/mitar_m
https://noc.social/@mitar