Move commentary on HTTP/3 frame and QUIC packetization #363
Conversation
LPardue
requested review from
rmarx,
marten-seemann and
lnicco
as code owners
| @@ -166,6 +166,16 @@ layer, and there isn't always a one to one mapping between HTTP and QUIC events. | |||
| The exchange of data between the HTTP and QUIC layer is logged via the | |||
| "stream_data_moved" and "datagram_data_moved" events in {{QLOG-QUIC}}. | |||
|
|
|||
| HTTP/3 frames are transmitted on QUIC streams, which allows them to span | |||
| multiple QUIC packets. Some implementations might send a single large frame, | |||
| rather than a sequence of smaller frames, in order to amortize frame header | |||
There was a problem hiding this comment.
Do we need to specify that this only applies to DATA frames?
There was a problem hiding this comment.
Ah I see the disjoint in my previous understanding of why the text was the way it was. In theory extension frames could do something similar but nothing concrete is in scope, only DATA benefits from that usage.
Frames can span packets and there might be some value is explaining that here. Let me try to tweak the text a bit more.
There was a problem hiding this comment.
I actually think the text might be fine as-is? There's also the case where you have the H3 frame headers in Packet N and the actual payload in Packet N+1 (say you have multiple streams being packed into the same packet for some reason), so not tying this down to DATA only seems fine.
If you really want, you could change it to Additionally, some implementations might send a single large frame, rather than a sequence of smaller frames..., so it's a bit clearer that the "single large frame" is a common occurrence of this issue but not the sole reason :)
Closes #280