Re: [quicwg/base-drafts] Allow removal of ACK blocks that are in flight (#917)

[Mike Bishop <[e-mail]>]

Can you help me understand the pathological case when something is ACK’d, omitted, then ACK’d again?  If the original ACK gets through, the sender should ignore the second one.  If the original ACK gets lost, it looks to the sender like a severely delayed packet and it will realize the retransmission was spurious (if it still has data), but might update the congestion controller with the un-loss event.  That doesn’t seem so bad at first glance.

From: ianswett [mailto:[e-mail]]
Sent: Sunday, November 12, 2017 4:42 PM
To: quicwg/base-drafts <[e-mail]>
Cc: Subscribed <[e-mail]>
Subject: Re: [quicwg/base-drafts] Allow removal of ACK blocks that are in flight (#917)


I think the existing text is mostly sufficient(though I realize it's a bit redundant at the moment, which I'll fix). I really do not advise some clever scheme where you send an ack block once and then then stop sending it for some arbitrary period of time and then start again, so I don't want to write text implying I'd recommend it.

There is still work to be done to add more discussion of when to send acknowledgements and what to include in them, but it was decided in Seattle that the transport draft should focus on the boundaries of what the receiver should do and the recovery draft can discuss the nuances and optimizations that are possible with considerations about their pros and cons. That work is tracked in #230<https://github.com/quicwg/base-drafts/issues/230>

"To limit receiver state or the size of ACK frames, a receiver MAY limit the number of ACK blocks it sends. A receiver can do this even without receiving acknowledgment of its ACK frames, with the knowledge this could cause the sender to unnecessarily retransmit some data. When this is necessary, the receiver SHOULD acknowledge newly received packets and stop acknowledging packets received in the past. Removing ack blocks before the peer has received them increases the chances the peer never received an acknowledgement, will declare the packet lost, and spuriously retransmit information."

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub<https://github.com/quicwg/base-drafts/issues/917#issuecomment-343722015>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AEE2hYPDupg70sZEpl27EPsg_D_5t6POks5s1q9ugaJpZM4Qaw74>.


-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/quicwg/base-drafts/issues/917#issuecomment-343740140