tcp congestion control algorithms and vista's new stack

[kamper]

So this link was posted in OS about upcoming vista changes: http://channel9.msdn.com/ShowPost.aspx?PostID=117938

In the middle of the video or so, they discuss new tweaks to the tcp congestion control algorithms. Just wondering if anybody knows much about this and can discuss specifics.

A couple things I'm curious about (I don't know anymore about this than my 3rd year networking course taught me):
-the algorithms are usually publicly discussed and agreed upon right? Where is the rest of the industry in regards to these changes?
-The guy mentioned that you'll see improvements between vista and a different machine if vista is receiving the majority of the data (eg downloading something). But don't most of the congestion control decisions happen on the sending side?

 

[Nothinman]

I don't know which algorithms Vista will use, but Linux currently supports 7:

Binary Increase Congestion (BIC)
TCP Westwood+
H-TCP
High Speed TCP
TCP-Hybla
TCP Vegas
Scalable TCP

2 of them have web pages in their description:

http://www.csc.ncsu.edu/faculty/rhee/export/bitcp/
http://www.icir.org/floyd/hstcp.html

 

[Smilin]

Although they have been, it's no requirement that they be publicly discussed and agreed upon. I think your concern may be something proprietary that breaks standard TCP/IP.

As long as a packets is properly sent there is no issue. When I say properly sent I mean: frame and sequence properly. Don't try using SACK or other options that weren't discussed in session setup. Don't try sending if the recieve window is full etc...

With the above in mind, you can make a great deal of changes to WHEN (or if) a packet is sent without breaking any rules.