Proggit Thread on Faster TCP Slow Starts

[dbrunton]

My little brother wrote a blog post, which I put into a reddit-ized headline (f-word, hyperbole, accusations, etc.), which actually turned into an interesting conversation. ON FRIGGING REDDIT. Wonders never cease.

[dbrunton]

Huh, for some reason, I can't see the comment I submitted with the post. Sorry if this shows up twice...

My little brother (I'm so proud!) wrote a blog post about faster TCP slow starts. He was sad nobody was reading it, so I redditized the headline (you know, f-word, hyperbole, baseless accusations, etc.) and submitted it to Reddit, where it actually turned into an interesting conversation about the actual topic of the post. Seriously. On Reddit.

Wonders never cease.

[aristus]

Credit should go to the Google "IW10" team, which did the hard work of testing this over years and (even harder) patiently convincing people one by one that this is a good idea.

I caution against fiddling with this number too much. Bandwidth-constrained users, and possibly mobile users, might be hurt with a too-large initcwnd.

http://code.google.com/speed/protocols/tcpm-IW10.html

[douglashunter]

Yes, tons of credit due there. In my post (http://bits.shutterstock.com/?p=74) I linked to Jonathan Corbet's coverage (http://lwn.net/Articles/427104/) of the topic on LWN, which is also a good jumping-off point for reading on the topic.

[rorrr]

SPDY

http://www.chromium.org/spdy/spdy-whitepaper