Open Bug 1528284 Opened 7 years ago Updated 3 years ago

Investigate our h2 rx granularity

Categories

(Core :: Networking: HTTP, enhancement, P5)

Product:
Core
Component:
Networking: HTTP
Type:
enhancement

Tracking

()

Tracking Flags:
Tracking Status
firefox66 --- affected
firefox67 --- affected

People

(Reporter: mayhemer, Unassigned)

Details

(Whiteboard: [necko-triaged])

I can often observe content coming in IMO too small chunks (0.5kB - 4kB) over h2. Each chunk adds some overhead.

Is this expected, actually good or something we may improve by changing something on our side? Like rwin size or its announcement or something?

How does other browsers behave? Is this server specific?

I promised to provide a log where this can be clearly demonstrated, so ni? myself.

Flags: needinfo?(honzab.moz)

(to get it off the triage list)

Assignee: nobody → honzab.moz
Flags: needinfo?(honzab.moz)
Priority: -- → P5
Whiteboard: [necko-triaged]

Dragana, could it be caused by us reading small amounts of the data from the socket? like reading only portions of control packets, nearly byte by byte. could it cause the kernel to send small range acks and potentially lead to some kind of disruption of the flow control?

Flags: needinfo?(dd.mozilla)

We should investigate this. Can you make a pcap?

Flags: needinfo?(dd.mozilla)
Assignee: honzab.moz → nobody
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.