1509827 - Document different classes of nsIClassOfService

Status: RESOLVED FIXED

(Core :: Networking, enhancement, P1)

Description

[Olli Pettay [:smaug][[e-mail]]]

Right now most of the classes in nsIClassOfService are undocumented and it is unclear how they relate to use of nsISupportsPriority.

Comment 1

[Honza Bambas (:mayhemer)]

Most of the classes specify a group from our h2 priority dependency tree:

translated to the h2 language:

https://searchfox.org/mozilla-central/rev/f997bd6bbfc4773e774fdb6cd010142370d186f9/netwerk/protocol/http/Http2Stream.cpp#1279-1305

and then used here:

https://searchfox.org/mozilla-central/rev/f997bd6bbfc4773e774fdb6cd010142370d186f9/netwerk/protocol/http/Http2Session.cpp#1089-1112

Comment 2

[Olli Pettay [:smaug][[e-mail]]]

This would be really useful for anyone trying to optimize page load.

Comment 3

[Honza Bambas (:mayhemer)]

Dragana, I'd like to ask for some details on the following 3 flags:

const unsigned long Follower = 1 << 1;
const unsigned long Speculative = 1 << 2;
const unsigned long Background = 1 << 3;

Thanks.

Comment 4

[Dragana Damjanovic [:dragana]]

(In reply to Honza Bambas (:mayhemer) from comment #3)

Dragana, I'd like to ask for some details on the following 3 flags:

const unsigned long Follower = 1 << 1;
const unsigned long Speculative = 1 << 2;
const unsigned long Background = 1 << 3;

Thanks.

It looks like we are not really setting this flags as we should. We should look into that.
Speculative - this is for speculative loads, something like prefetch, preload.
Follower - should correspond to not having NS_HTTP_LOAD_AS_BLOCKING nor NS_HTTP_LOAD_UNBLOCKED flag (I think we do not set this right).
Background - currently is only used here https://searchfox.org/mozilla-central/source/dom/base/Navigator.cpp#1189

It would be good to document this better and they to use this flags more in the code.

Comment 5

[Olli Pettay [:smaug][[e-mail]]]

Does speculative mean also the loads initiated by HTML parser when it does its speculative loading stuff?

Comment 6

[Honza Bambas (:mayhemer)]

(In reply to Olli Pettay [:smaug] from comment #5)

Does speculative mean also the loads initiated by HTML parser when it does its speculative loading stuff?

looks unused: https://searchfox.org/mozilla-central/search?q=symbol:E_%3CT_30d3568e12e7221e%3E_Speculative%2CE_%3CT_585b7a22%3E_Speculative%2CE_%3CT_e2e10413af089363%3E_Speculative%2CE_%3CT_ef70bfffab818152%3E_Speculative&redirect=false

Comment 7

[Boris Zbarsky [:bzbarsky]]

Note that if we do use the speculative thing for the HTML parser speculative loads we need a way to "switch" to non-speculative once the real load tries to coalesce with the speculative load.

Comment 8

[Olli Pettay [:smaug][[e-mail]]]

Honza, could this be prioritized a bit?

Comment 9

[Honza Bambas (:mayhemer)]

Yup, will look at this ASAP.

Comment 10

[Honza Bambas (:mayhemer)]

Attachment: Bug 1509827 - Document nsIClassOfService flags, r=dragana

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/d37f02809ffd
Document nsIClassOfService flags, r=dragana

Comment 12

[Stefan Hindli [:stefan_hindli]]

https://hg.mozilla.org/mozilla-central/rev/d37f02809ffd