Re: [Cake] memory

[Sebastian Moeller <moeller0@gmx.de>]

Hi Toke,


On Oct 29, 2015, at 11:14 , Toke Høiland-Jørgensen <toke@toke.dk> wrote:

> Sebastian Moeller <moeller0@gmx.de> writes:
> 
>> Fair enough ;) Let’s get the soldering iron out and see whether I can
>> replace the phi with a 10Mbit one ;), mmmh, maybe I should run 10BASE5
>> and do away with all that modernist “hub” and “switch” witch craft
>> while I am at it Joking aside, I believe the issue only occurred on
>> loooong RTTs so it will only be a minor issue in real life, but still
>> it would be nice if cake would not introduce the possibility for
>> remote parties to make my router OOM/reboot ;). I might be odd, but I
>> prefer a hit in throughput to a forced reboot any day ;)
> 
> The "1 second RTT" part was a key part in that... If the RTT is set to
> 100ms (the default), you'll only get a limit of 12.5 MB…

	Yes, while trying to be funny I glossed over that part a bit (the loooong RTTs part was aiming at that), sorry. But really I am wonder what it is that makes my position that cake should expose and honor a packet limit for its queueing so controversial; I am totally happy with cake automatically selecting a good default automatically using all information it has available, like configured bandwidth and RTT, all I want to avoid the situation where cake divined something that is totally consistent within the information it has available, but is obviously wrong given additional information out of cake’s control (like total memory size or even maximum memory size the user is willing to put aside for cake). I think cake’s goal of being simple to use is commendable, but we should make sure we do not accidentally end up with cake being simplistic… (rather make the normal things easy/simple and complex/crazy things possible ;) )

Best Regards
	Sebastian


> 
> Now, there is still an issue to be resolved in "unlimited" mode. But
> that's a separate issue.
> 
> -Toke