Trustless Gateway format parameter should take priority over accept header?

[achingbrain]

The Trustless Gateway spec says of the format parameter:

When both the Accept header and format parameter are present, a specific Accept value (e.g., application/vnd.ipld.raw) SHOULD take precedence over format.

As written this prevents browsers from using the format parameter to control how a resource is returned.

For example if I request a block that can be rendered as HTML is requested as raw using the format parameter an Accept header with specific (e.g. non-wildcard) values is always sent:

..so following the spec the block should be rendered as HTML and the format param ignored.

I think this is the wrong way round - e.g. the format parameter should always take priority over the Accept header if it is present.

[lidel]

iirc the reason for this was historical (Project Rhea or earlier): in the past Accept header could include more information than format (requesting specific block order in car etc), it could also support fallbacks (e.g. give me CAR but if not, fallback to raw block) to avoid extra roundtrips.

i think these days we have means of requesting that via url query params as well, thus we could make the change, but it would be a breaking change in https://github.com/ipfs/gateway-conformance, so requires IPIP to document and evaluate the change properly