Re: RenoTcpAgent::dupack_action()

[Sally Floyd <floyd@ee.lbl.gov>]

>If anyone is familiar with the TCP Reno code,
>I was wondering why the function dupack_action() does not perform a fast 
>retransmit when bug_fix_=1 and ecn_=0...  it seems like this forces the use of
>ecn_ (Explicit Congestion Notification) which may not be desirable for all 
>simulations...   

The motivation for bug_fix_ is discussed in the following note:

Floyd, S.,
<b><a href="ftp://ftp.ee.lbl.gov/papers/fastretrans.ps">
TCP and Successive Fast Retransmits.</a></b>
Technical report, October 1994.

Its use with Reno TCP is illustrated in the validation test
"./test-all-newreno", in the two tests 
  ../../ns test-suite-newreno.tcl reno2
and
  ../../ns test-suite-newreno.tcl reno2_bugfix
These are also shown in the following:

Floyd, S.,
August 24-28,
Short presentation on <b>Revisions to RFC 2001</b> 
(<a href="ftp://ftp.ee.lbl.gov/talks/sf-tcpimpl-aug98.ps">postscript</a>,
<a href="ftp://ftp.ee.lbl.gov/talks/sf-tcpimpl-aug98.pdf">PDF</a>).

It is not about favoring ECN.   It is about avoiding multiple
fast retransmits from a single window of data.

- Sally