It would be awesome if Tomato could support the new Codel algorithm that goes a long way toward fixing buffer bloat problems. OpenWRT has it in it's mainline but it requires an upgrade to the 3.3 branch of the Linux Kernel.
Date: 29 Jun 2012 22:00
Number of posts: 12
RSS: New posts
The problem AFAIK is not kernel support but the Broadcom wireless drivers which are still proprietary, and are designed to work with kernel 2.6.22 only. Which is why Tomato is still running with a kernel from 2007, albeit with some backports from newer kernels. Even the DD-WRT guys who have special agreements with Broadcom cannot get higher than 2.6.24.
The only thing we can do is to wait for Broadcom to either open source all their drivers (very unlikely) or release a new platform SDK based off a newer kernel, like what happened when they switched from K2.4.20 to K2.6.22 with the advent of wireless-n. The newer drivers happened to also support g-only radios so WRT54Gs could then run K26 with everything working. Maybe Wireless-AC routers will bring that change ?
Other chipsets with open source drivers for all the components (like Atheros) do not have this limitation.
An issue that is totally insignificant to puny home routers and puny connection speed, but some people create that issue in their mind an waste time fixing it on a home router.
@kaba is incorrect - the problem is almost *exclusively* with home routers and puny connection speeds.
See the wikipedia page on bufferbloat (I cannot post links) for more information.
Wall of text with no quantified metric is horse shit. All that I read are "feeling" and "hunches". Stop the insanity of modifying buffer for sake of modifying to satisfy "feelings". Show proper benchmarks that will affect real use, no academic theories.