prevent scheduled ACKs from getting cancelled

[Lanius-collaris]

Description

I hope it works. 😇

Reference issue

Fixes #461

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.16%. Comparing base (e3a1986) to head (b2c3092).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #462      +/-   ##
==========================================
+ Coverage   84.13%   84.16%   +0.03%     
==========================================
  Files          54       54              
  Lines        4287     4283       -4     
==========================================
- Hits         3607     3605       -2     
+ Misses        494      493       -1     
+ Partials      186      185       -1     

Flag	Coverage Δ
go	84.16% <100.00%> (+0.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[JoTurk]

This will have extra cost of sending some ACKs slightly earlier in a few bursty cases under high throughput. But i think this is fine, we can benchmark it later.
We'll need a unit test for this behavior before we can merge it.

[Lanius-collaris]

This will have extra cost of sending some ACKs slightly earlier in a few bursty cases under high throughput.

Two controllers in cc-alg branch have CanSend() method to tell the caller when to send packets.
https://github.com/Lanius-collaris/pion-sctp/blob/5af9dde48b75c9cc3c6bbb00310976daf4f57e65/cc_alg/frcc.go#L700
https://github.com/Lanius-collaris/pion-sctp/blob/5af9dde48b75c9cc3c6bbb00310976daf4f57e65/cc_alg/reno.go#L71
Do you want to merge it first?

Interleaving nonblocking Read() with "blocking" Read() ( e.g. net.Conn.Read ) can reduce ACKs too.

[Lanius-collaris]

We'll need a unit test for this behavior before we can merge it.

I don't know how to write unit tests for multi-threaded code.

This will have extra cost of sending some ACKs slightly earlier in a few bursty cases under high throughput. But i think this is fine, we can benchmark it later. We'll need a unit test for this behavior before we can merge it.

If you have time, could you benchmark my reno.go? I didn't change the default algorithm in cc-alg branch.
https://github.com/Lanius-collaris/pion-sctp/blob/5af9dde48b75c9cc3c6bbb00310976daf4f57e65/association.go#L633-L679

Also, Reno-style CCAs' AIMD behavior (additive-increase, multiplicative-decrease) make them unable to utilize most of available bandwidth when the bandwidth × delay product is large or the packet loss rate isn't small enough, so don't worry.
https://intronetworks.cs.luc.edu/current/html/dynamicsB.html#the-high-bandwidth-tcp-problem
https://intronetworks.cs.luc.edu/current/html/dynamicsB.html#the-lossy-link-tcp-problem