Examples have diverged from the definitions in out-of-order packets section

[aochagavia]

I'm implementing the ACK frequency draft and it looks like the examples have diverged from the definitions in the section titled "Response to Out-of-Order Packets".

When looking at the text, the definition of "Unreported Missing" says they are "Packets with packet numbers between the Largest Unacked and Largest Acked that have not yet been received" (emphasis mine). When looking at the examples, however, it looks like the definition is slightly modified to include all missing packets lower than Largest Unacked (i.e. even when they are lower than Largest Acked).

Below I have worked out the provided examples, to show how the mismatch looks in practice. Notice how, in both examples, an ACK is sent in the last line even though there are no unreported missing packets between Largest Unacked and Largest Acked.

Which one is correct? The text or the examples?

Example 1

reordering_threshold = 3

largest_unacked = 0, largest_acked = 0, unreported_missing = [], gap = 0
Receive 1
largest_unacked = 1, largest_acked = 0, unreported_missing = [], gap = 0
Receive 3
largest_unacked = 3, largest_acked = 0, unreported_missing = [2], gap = 1
Receive 4
largest_unacked = 4, largest_acked = 0, unreported_missing = [2], gap = 2
Receive 5
largest_unacked = 5, largest_acked = 0, unreported_missing = [2], gap = 3
===> Send ACK, because gap >= reordering_threshold
largest_unacked = 5, largest_acked = 5, unreported_missing = [], gap = 0
Receive 8
largest_unacked = 8, largest_acked = 5, unreported_missing = [6, 7], gap = 2
Receive 9
largest_unacked = 9, largest_acked = 5, unreported_missing = [6, 7], gap = 3
===> Send ACK, because gap >= reordering_threshold
largest_unacked = 9, largest_acked = 9, unreported_missing = [], gap = 0
Receive 10
largest_unacked = 10, largest_acked = 9, unreported_missing = [], gap = 0
===> The draft says: "Send ACK because of 7", which means it considers 7 to be unreported missing!

Example 2

reordering_threshold = 5

largest_unacked = 0, largest_acked = 0, unreported_missing = [], gap = 0
Receive 1
largest_unacked = 1, largest_acked = 0, unreported_missing = [], gap = 0
Receive 3
largest_unacked = 3, largest_acked = 0, unreported_missing = [2], gap = 1
Receive 5
largest_unacked = 5, largest_acked = 0, unreported_missing = [2, 4], gap = 3
Receive 6
largest_unacked = 6, largest_acked = 0, unreported_missing = [2, 4], gap = 4
Receive 7
largest_unacked = 7, largest_acked = 0, unreported_missing = [2, 4], gap = 5
===> Send ACK, because gap >= reordering_threshold
largest_unacked = 7, largest_acked = 7, unreported_missing = [], gap = 0
Receive 8
largest_unacked = 8, largest_acked = 7, unreported_missing = [], gap = 0
Receive 9
largest_unacked = 9, largest_acked = 7, unreported_missing = [], gap = 0
===> The draft says: "Send ACK because of 4", which means it considers 4 to be unreported missing!

[mirjak]

Thanks for flagging this. It has been challenging to get the text write that's why we included examples. We will revisit the text again and try to get it straight!

[huitema]

I think this is somewhat related to #213.

[ianswett]

Thanks for filing this!

I believe #222, which fixed #213, caused the text to line up with the examples. @aochagavia Can you verify that and if so close this issue?

[aochagavia]

Thanks for fixing it :)

I don't have the bandwith right now to check, so if you think the issue is fixed feel free to close the issue. I'll also give a heads up to the folks developing Quinn, so if there is any ambiguity left they'll probably find it.

[ianswett]

Thanks!