> On 21 Jun 2025, at 2:19 PM, David Brownlee <a...@absd.org> wrote:
> 
> On Thu, 19 Jun 2025 at 01:25, Emmanuel Nyarko <emmankoko...@gmail.com> wrote:
>> 
>>> On 18 Jun 2025, at 6:54 PM, Emmanuel Dreyfus <m...@netbsd.org> wrote:
>>> 
>>> On Wed, Jun 18, 2025 at 03:11:30PM +0000, Emmanuel Nyarko wrote:
>>>> 1. Remove all the existing disciplines (openBSD has taken this step 
>>>> already)
>>> 
>>> That will break the setups of long-time users.
>>> 
>>> But if fq-codel is able to replace all existing disciplines, perhaps
>>> you could keep the existing configuration keywords for other disciplines,
>>> and use them to configure fq-codel to emulate the old disciplines.
>> 
>> sadly, that doesn’t look possible. Fq-codel is just three params(IRC) and 
>> the rest are a lot for a single queue.
> 
> If there really no useful setups which fq-codel could replace?
> Including "I need to simulate packet loss and odd bandwidth behaviour
> to stress my other systems"?
> 
> If so, I think there are two aspects to "replace everything with fq-codel":
> 
> - Can existing configuration syntax recognised and mapped to fq-codel
> (likely with a run-time warning), which could be done in one of two
> ways
>  - Map each existing discipline to a given default fq-codel config
>  - As above, but take some subset of the params to adjust fq-codel
> (very much only a subset)
> 
> Given the above, would there be any significant functional regressions
> in any existing ALTQ installations were converted to fq-codel?
> 
> Thanks
> 
> David

https://undeadly.org/cgi?action=article;sid=20140419151959

this is what OpenBSD has done, i looked into it kind of was convincing me 
enough. 


But then, i can actually leave the old ones and add fq-codel separtely on a 
different design so we can easily modularize and also become easily extensible 
in other modular systems.(maybe would be a hard one) but I would look into it.

Emmanuel





Reply via email to