In message <200007100349.XAA13570@ghost.whirlpool.com>, Andrew Gillham writes:
>John Hawkinson writes:
>> | This should all be part of QoS under NetBSD.
>> 
>> I don't really see what QOS has to do with this discussion.
>> That sounds very seperate to me.
>
>Hmm, what is the point of preventing a DoS if you aren't interested
>in QoS?

The same point as anything else if you're not interestewd in QoS.
I don't understand your assertion.

Preventing a denial of service is useful so you can have a service.
Adding QoS is a giant can of worms, the involing politics of how
to run and use networks, and hosts cooperating with routers, etc., etc.


>Ensuring that resources are shared appropriately and can't be
>"hogged" by one user (e.g. a DoS) are fundamental to QoS, IMHO.

This hardly means that QOS is fundamental to people who want to 
prevent DOS.

>Certainly "rate limiting" and "queueing" are very different from a
>policy perspective, but the underlying software is more or less the
>same.  e.g. I might want to implement Weighted Fair Queuing on a WAN
>link, while also limiting specific traffic (like ICMP) to a certain
>rate.  Or I may want to reserve a certain amount of bandwidth for a
>specific flow.

It is my judgement that any discussion of QoS is really totally
irrelevent in a release-focussed (i.e. 1.5-focussed) discussion.
We're talking about fixing something that is broken, not adding
complicated hairy new mechanisms.

Such a discussion is probably reasonable, but it is not this one.
I would like to stay focussed on how we're fixing ICMP rate-limitting
for v4 (and v6) here.

I'd ask you to use a different thread for QoS discussions.
As Itojun points out, integrating ALTQ for 1.5 is not reasonable.

--jhawk