Re: New Version Notification for draft-kazuho-httpbis-priority-00.txt

[Lucas Pardue <[e-mail]>]

Hi Tom,

Thanks for your sharing your thoughts. I'd like to respond in more direct
detail but fear I might tread on the conversation toes.

To talk around a general topic of discussion, I've been noodling the
philosophy of priorities today. One thing that emerges to me is that a
client parsing a document has a _vague idea_ what the resource is, it can
determine some information from link relation, element, file extension etc
but it is only a partial picture. Clients typically assign their H2
priorities at this stage, based on the intended usage. What they can't know
exactly is the selected representation [1], which is determined by the
server. For example, whether a JPEG image is encoded as progressive or
baseline. The nature of the selected representation may affect how the
client would use it, and it is challenging for a server to reflect this
back into all the flavours of H2 priority tree that exist. And although a
client could perhaps apply reprioritisation to the in-flight response
(after it learns information), chances are that it will happen too late.

Since the selected representation is chosen during the lifetime of an
end-to-end connection, it is difficult to define a static signal that can
work for all. So my opinion is that the proposal is less about one side
running roughsod over the other, but all sides attempting to paint a fuller
picture using a common set of accessible terms. Of course this might go
askew, your example of a background priority getting "upgraded" has come up
- there is space in the spec for this to be addressed IMO.

Kinds regards
Lucas

[1] https://tools.ietf.org/html/rfc7231#section-3