Re: HTTP/2 Priorities Proposal

On Wed, Apr 9, 2014 at 10:20 AM, Jeff Pinner <jpinner@twitter.com> wrote:

>
>> The one dependency tree is easily collapsed to multiple trees in usual
>> situation. For example, if 5 streams depend on stream A, and A is closed,
>> those 5 streams become root for each subtree. And changing weight of them
>> requires 5 priority frames.
>>
> One way to consider this: stream A, as the root of the its dependency
> tree, had a weighted edge from stream 0 (this is equivalent to saying it
> was in a group with some weight). When A closes, it's 5 dependents each now
> become roots of their own subtrees, so each has a edge from stream 0. The
> question posed is on stream closure, what should those weights be? If the
> weight of the edge to A is distributed to the dependents in proportion to
> their weights on A, no changing of weights is required and we maintain
> equivalence with the draft-11 proposal.
>

My concern is chaining weight of streams in general, which happens *after*
direct descendants of A becomes root (linked from stream 0) and A's weight
was distributed.

In -11, after closure of A, all descendants are still in the same group, we
can take one of the root and specify weight. It is done with 1 PRIORITY.
 In the proposed plan, descendants are now root of its own subtree,
disconnected each other. To change all weights, we have to issue PRIORITY
frames for each root.

The well known example of this usecase is tab browsing.  People change tab.
Browser wants to change weights of streams, grouped by a tab.

Best regards,
Tatsuhiro Tsujikawa