Re: AltQ

["Thierry Deval" <Thierry.Deval@Computer.Org>]

Hi,

On Tuesday, January 25, 2000 2:38 AM, Arkan mailto:ark@eltex.ru wrote:
> nuqneH,
>
> Hope you mean ipf?

Yes an No...

Ipf would certainly be the best interface for setting up the queuing
rules, but afaik it is limited to the IP protocol.
While altq and dummynet are all still IPv4 queuing tools, I see no
reason to not extend the scope of such tool to other protocols (esp.
IPv6, but also any other).
Moreover, it should not be limited to rule on physical interfaces (eth,
atm),  but on any network facility (ppp, gif) as does certainly ipf.
Which is why the altq approach is wrong, isn't it ?

Then we should devise whether ipf would embrace any protocol (a la
ifconfig) and move it away from netinet (is it possible ?), or add some
other filter ?

Maybe yes, we should start on ipf, though...

> Thierry Deval <TDeval@PrimeOBJ.COM> said :
>
> > OK, my best best would be to base alt-q on bpf, or maybe just under
it.
> >
> > I'll look at all this HUGE networking code, and see were AltQ could
fit.
> > Any suggestion from you gurus are welcome. And when I come back with
something,
> > there will certainly be a need for deep cleaning ;-)
          ^ ^ ^ ^
          | | | |

To help me, could someone draw a summary on how the different parts of
the networking code fit together ? ( from layer 1 <-- ... -->
application layer )