1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
|
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "nsISupports.idl"
/**
* nsIClassOfService.idl
*
* Used to express class dependencies and characteristics - complimentary to
* nsISupportsPriority which is used to express weight
*
* Channels that implement this interface may make use of this
* information in different ways.
*/
// convenience class for passing around the class of service
%{C++
namespace mozilla::net {
class ClassOfService;
}
%}
native ClassOfService(mozilla::net::ClassOfService);
[scriptable, builtinclass, uuid(1ccb58ec-5e07-4cf9-a30d-ac5490d23b41)]
interface nsIClassOfService : nsISupports
{
attribute unsigned long classFlags;
attribute boolean incremental;
void clearClassFlags(in unsigned long flags);
void addClassFlags(in unsigned long flags);
[noscript] void setClassOfService(in ClassOfService s);
// All these flags have a (de)prioritization effect.
// In the HTTP/1 world the priority is considered for all requests inside a so
// called 'Request Context' which is a context common to all sub-resources
// belonging to a single top level window (RequestContextService). Requests
// marked with the Leader flag are blocking (preventing from being sent to the
// server) all other resource loads except those marked with the Unblocked
// flag. Other classes run in parallel - neither being blocked no ;r blocking.
// The Leader flag is used only for <head> blocking resources (sync and
// defer javascript resources and stylesheets.) Purpose is to deliver these
// first-paint and domcontentloaded blocking resources as soon as possbile.
// In the HTTP/2 world it's different. Priorities are done only per HTTP/2
// session, normally we have one session per one origin (including origin
// attributes!) Requests are dispatched (sent) immediately on an HTTP/2
// session. Each session has artificial streams (groups) relating to the class
// of service flags (Leader, Other, Background, Speculative, Follower,
// UrgentStart), each such a stream is given a different weight (only way to
// give a stream a priority in HTTP/2) reflecting the desired request group
// priority. Actual request streams are then dependent on these artificial
// streams (groups). nsISupportsPriority of each request is passed as a weight
// on the HTTP/2 stream to prioritize streams in the same group. A stream can
// also be dependent on other stream. We have dependency of Followers on
// Leaders, hence everything set the Follower flag should be processed by the
// server after Leaders. Same for Speculative being dependent on Background. The
// tree is created and used here:
// https://searchfox.org/mozilla-central/rev/cc280c4be94ff8cf64a27cc9b3d6831ffa49fa45/netwerk/protocol/http/Http2Session.cpp#1053-1070
// https://searchfox.org/mozilla-central/rev/cc280c4be94ff8cf64a27cc9b3d6831ffa49fa45/netwerk/protocol/http/Http2Stream.cpp#1338
// For detailed description of how HTTP/2 server should handle the priorities
// and dependencies see:
// https://developers.google.com/web/fundamentals/performance/http2/#stream_prioritization
// Please note that the dependecies and weights we are sending to the server
// are only suggestions, the server may completely ignore it.
// Leaders (should) block all other resources except Unblocked. This flag
// also priortizes HTTP cache reading queue by blocking all other cache
// requests.
const unsigned long Leader = 1 << 0;
// The Follower flag is currently unused!
const unsigned long Follower = 1 << 1;
// The Speculative flag is currently unused!
const unsigned long Speculative = 1 << 2;
// The Background flag is currently only used for Beacon.
const unsigned long Background = 1 << 3;
// Requests marked with this flag are not blocked by Leaders. This is mainly
// used for probing-like XMLHttpRequests that may block delivery of head
// blocking resources, e.g. CSS files tailored for the UA.
const unsigned long Unblocked = 1 << 4;
// Throttleable flag allows response throttling of the resource load. Note
// that this functionality is currently disabled.
const unsigned long Throttleable = 1 << 5;
// UrgentStart makes the request temporarily extend HTTP/1 connection
// parallelism limits. Used mainly for navigational requests (top level html)
// and any request considered coming from a user interaction to make reaction
// of the browser as fast as possible and not blocked.
const unsigned long UrgentStart = 1 << 6;
// Specifically disables throttling under any circumstances, used for media
// responses mainly.
const unsigned long DontThrottle = 1 << 7;
// Enforce tailing on this load; any of Leader, Unblocked, UrgentStart,
// TailForbidden overrule this flag (disable tailing.)
const unsigned long Tail = 1 << 8;
// Tailing may be engaged regardless if the load is marked Unblocked when some
// other conditions are met later, like when the load is found to be a
// tracker.
const unsigned long TailAllowed = 1 << 9;
// Tailing not allowed under any circumstances or combination of flags.
const unsigned long TailForbidden = 1 << 10;
};
|
