Bug 2076

Summary:	MinRTO Attribute is not actually used.
Product:	ns-3	Reporter:	Tommaso Pecorella <tommaso.pecorella>
Component:	tcp	Assignee:	Tommaso Pecorella <tommaso.pecorella>
Status:	RESOLVED FIXED
Severity:	minor	CC:	natale.patriciello, ns-bugs, tomh
Priority:	P5
Version:	ns-3-dev
Hardware:	All
OS:	All
Attachments:	patch

Description Tommaso Pecorella 2015-03-12 14:23:46 UTC

The glitch has been reported by Jonny Omer here:
https://groups.google.com/forum/#!topic/ns-3-users/78ThKfmPo18

Indeed, TcpSocketBase has an Attribute named "MinRTO", but it's not used once in the code.
Either we remove the attribute (not a good idea) or we use it.

Moreover, in the actual code the MinRTO is hardcoded to 1 second (RFC 6298), but MinRTO attribute is set to 0.2 seconds.
In order to prevent a sudden change in TCP behaviour, it's a good idea to change the MinRTO default value to 1 second.

Comment 1 Tommaso Pecorella 2015-03-12 14:30:23 UTC

Created attachment 1988 [details]
patch

Comment 2 natale.patriciello 2015-03-13 05:28:52 UTC

(In reply to Tommaso Pecorella from comment #0)
> 
> Moreover, in the actual code the MinRTO is hardcoded to 1 second (RFC 6298),
> but MinRTO attribute is set to 0.2 seconds.
> In order to prevent a sudden change in TCP behaviour, it's a good idea to
> change the MinRTO default value to 1 second.


I think 0.2 seconds is taken from Linux.

I think it is a good idea to state "we follow RFC etc..., for a Linux-compatible behavior, please select 0.2 etcetc" somewhere in the documentation. What do you think?

Comment 3 Tom Henderson 2015-03-14 15:58:14 UTC

I'm fine with this patch; I'm neutral about changing the attribute default, but since we seem to have been running at 1 second anyway due to the bug, we probably don't need to note this as "changed behavior" in the release notes.

Comment 4 Tommaso Pecorella 2015-03-15 03:45:33 UTC

changeset:   11231:62ce11a18acd