Re: HTTP/3 with strict priorities (PR #2700)

[Ian Swett <[e-mail]>]

On Fri, May 17, 2019 at 10:56 AM Dmitri Tikhonov <
[e-mail]> wrote:

> On Wed, May 15, 2019 at 05:51:27PM -0400, Ian Swett wrote:
> > > 1. What is a _strict priority_?
> > >
> > >     There is no definition of strict priority that I could find
> > >     in the diff.  Do strict priorities work outside of the
> > >     dependency tree?  For example, which stream has higher priority
> > >     in the following figure?
> > >
> --- 8< --- 8< --- 8< ---
> >
> > Strict priorities are only relative to other siblings under the same
> > placeholder/root.  I'll clarify that.  Currently the text says "Higher
> > priority elements have all available data sent before elements of lower
> > priority."
>
> I see, so this is two levels of prioritization, with weights (if present)
> serving to determine how resources are allocated to streams that have the
> same priority.  This feature is analogous to priority and concurrency in
> Patrick Meenan's proposal. [1]
>
> > > 3. What problem do all-or-nothing resources solve?
> > >
> > >     " When [weight is] absent, indicates the resource should be
> > >     " delivered all at once or not at all.
> > >
> > >     This is new.  I don't believe I've seen it in other prioritization
> > >     proposals.  Adding this dimension increases complexity.  What
> > >     problem do these all-or-nothing resources solve?
> > >
> >
> > This is an effort to incorporate Patrick Meenan's Round Robin Sequential
> > and Round Robin Interleaved proposal, as outlined on slide 49 here:
> >
> https://github.com/HTTPWorkshop/workshop2019/blob/master/talks/pardue-jones-priorities.pdf
> > or
> > here: http3-prioritization-proposal
> > <
> https://github.com/pmeenan/http3-prioritization-proposal/blob/master/README.md
> >
> >
> > I'm sure Patrick could rewrite this text more clearly, and I think
> > Patrick's explanation for the motivation is solid.
>
> I could not find the "all at once or not at all" concept mentioned
> in these documents, nor did I succeed in mapping the "not at all"
> part to anything in Patrick's proposal.  Where can I find his
> explanation for the motivation?
>

I'm referencing "Sequential" and "Interleaved" on slide 49.

I added this text after your original review, which hopefully is clearer:
"Higher priority elements have all available data sent before elements of
lower
priority.  When multiple elements have the same priority, if a weight is
specified, elements with weights are interleaved.  If a weight is 0, either
because it is the default weight or it was not specified, then elements are
delivered sequentially and all available data from one element is sent
before
any data from the next.  If a given priority level has some elements with
weights and some without, the two groups share the available bandwidth
equally."


>   - Dmitri.
>
> 1. https://github.com/pmeenan/http3-prioritization-proposal
>
>