[OpenAFS] Network becomes terribly slow when cache manager flushes updates over xDSL

Carson Gaspar carson@taltos.org
Wed, 08 Jul 2009 10:36:26 -0700


Mattias Pantzare wrote:
> On Wed, Jul 8, 2009 at 19:06, Carson Gaspar<carson@taltos.org> wrote:
>> Sean O'Malley wrote:
>>
>>> -icmp shouldn't be used. Some BOFH block/drop icmp.
>> ...
>>> -is there a standard way to do UDP mtu link detection on any platform
>>> already?
>> The standard is ICMP WOULD FRAGMENT. People who block that don't deserve
>> functional apps. Don't cater to them.
> 
> The end user seldom have control over all firewalls...
> 
> There are two goos solutions to this problem, one is to start using
> TCP for AFS as these problems are solved in good TCP implementations.
> 
> The other solution is to look at how a good TCP implementation handles
> path mtu discovery, including black hole detection (when the wold
> fragment icmp packets are dropped) and copy that.

While I pity users of broken networks, any effort expended beyond "turn off PMTU 
discovery" to work around intentionally broken networks is completely wasted. 
And I'd argue that even _that_ is a bad idea that just enables stupid people to 
break their networks.

-- 
Carson