-
Notifications
You must be signed in to change notification settings - Fork 14
segments_acked rather than bytes_acked; cwnd in segments #67
Comments
segments_acked is used throughout the document to match the linux code. Section 4.1.2 (https://datatracker.ietf.org/doc/html/draft-ietf-tcpm-rfc8312bis-02#section-4.1.2) defines the units for all variables including
|
@nsdyoshi does this address your comment? |
Ok. I see the definition in cwnd. I'm sorry, I overlooked. |
Yes, implementations can use bytes_acked. Does the below text suffice:
|
I think it would be better to mention using bytes_acked is acceptable for implementations explicitly. BTW, when a segments_acked implementation sent a 1000 byte packet and only 500 bytes were acked, does it use fractions or use some other methods? |
If the size of each segment = 1 MSS, then segments_acked = bytes_acked / 1 MSS.
Good question, is that being done for new Reno or other congestion controllers? I looked at a few RFCs and didn't find any mention of using SACK bytes for congestion window. Do you know if any document has guidance on this? |
Right. It's described in RFC6582 and RFC6675. |
If I were to use segments_acked as an implementor, I would just do bytes_acked / 1 MSS, whatever may be the size of the segment sent or ACKed. For example if MSS = 1448 and 1000 byte segment was sent and acked, then segments_acked = 1000/1448 and if 500 bytes were acked, then segments_acked = 500/1448. If OTOH I was using bytes_acked in my implementation, then all one needs to do multiple or divide by 1MSS wherever necessary. |
OK. In this case, I think it would be better to mention segments_acked is float number and can be less than 1. |
ok, please review #77. |
Yoshi said:
The text was updated successfully, but these errors were encountered: