Re: Priority signals in HTTP/2

On Fri, 18 Dec 2020 at 03:26, Martin Thomson <mt@lowentropy.net> wrote:

> Now that we're official, I'd like to get started on some of the more
> difficult pieces: priority.
> I wrote up a proposed plan for dealing with the priority signaling in RFC
> 7540 here:
> https://github.com/httpwg/http2-spec/issues/773#issuecomment-725285414
> What do people think about this general approach (quoting):
>
> 1.    Cut the section with the detailed explanation and replace it with
> some explanatory text


OK


> about the importance of having prioritization in a multiplexed transport
> (though not necessarily signaling thereof).
>

Lost me here!    I'm just not sure that it is universally true that
priorities are needed in a multiplexed transport.   I'm sure there are some
circumstances that need them, but I am far from convinced that they are
generally required nor that the complexity of implementation is justified
in many of the cases that they might be conceptually useful.


> 2.    Explain that this document used to define a priority signaling
> scheme, but that the one that was defined was unsuccessful.
> 3.    In that section recommend against using the priority signaling
> scheme from RFC 7540, but explain that other endpoints that you talk to
> might send these frames.
>

OK - I'm back onboard with 2. and 3.

4.    Point to the new signaling system if that is mature enough (and
> probably even if it is not).
>

I don't think that any new priority signalling system should be linked or
recommended unless it has been proven to be of general benefit.
However, I'd not be opposed to defining a general signalling system that
would allow consenting endpoints to exchange arbitrary signals. Such a
system could be used as the basis for a priority conversation if/when
something is agreed, but would be ignorable by non consenting/participating
endpoints.  It would also better allow for experimenting with alternate
solutions and optimizations (eg perhaps some extension to flow control
would be better than priorities).


> 5.    Suggest, but not recommend, that servers pay attention to priority
> signals using these old mechanisms in the absence of other signals. This
> might take a little careful wording, but it's probably enough to caution
> that ignoring signals might degrade performance and some signals might be
> better than none. Point to RFC 7540 for the details of how the old scheme
> works.
>

As a server implementor (jetty), we have thought long and hard about any
way we can realistically implement priorities in a way that would not hurt
our general overall performance for non prioritized handling.  It may be
true that ignoring priorities MIGHT degrade performance, but I think it is
equally true that implementing priorities MAY itself degrade performance.
We've deployed HTTP/2 in some very large scale high performance
situations and our lack of priorities has not been an issue.

So the careful wording should say that the RFC7540 priorities may help
performance in some circumstances, but that it may also be valid to ignore
them. Ie it should say "your mileage may vary"!



> 6.    Leave the definition of the frame type intact, but cut the
> definitions right down to just the necessary pieces sizes, names, and any
> mandatory processing (error handling, sizes, ranges, the hard stuff). This
> should ensure that people are able to do things like generate the right
> error codes and skip over the HEADERS frame pieces properly.
> 7.    Leave text about when PRIORITY can be sent in the state machine.
> This is important for interoperating with endpoints that generate the frame.
>

Definitely need to keep this for interoperability.  Some endpoints will
send PRIORITY frames for the foreseeable future, so we can't pretend they
don't exist, even if we ultimately ignore them.


-- 
Greg Wilkins <gregw@webtide.com> CTO http://webtide.com