Pekka,
Mostly agree, except with the final paragraph where you say that
the parameters for token bucket, timer, etc. mechanisms don't
matter that much. ICMPv6 says that errors should include up to
the IPv6 min-MTU bytes of the packet that caused the error. That's
1280 bytes, but on some long, thin links the store-and-forward time
for such a large packet can be on the order of 1 second! (Remember
also the asymmetric path case I referred to in a previous message.)
So, it seems to me that it does matter that suitible values are chosen
for the parameters - am I wrong?
Fred
On Wed, 7 Jan 2004, Brian Haberman wrote:
> > I think that would appropriately encourage implementation of
> > the Token-bucket method without invalidating existing
> > implementations that use one of the others.
>
> I agree with the sentiment here. Changes to this document should
> not affect backwards compatability. So, I would be opposed to
> making the proposed change unless we can determine that noone has
> implemented the timer-based or bandwidth-based mechanisms.
Timer-based methods have, unfortunately, been implemented and
deploying. IMHO, those should definitely be changed, but I have some
sympathy for those who see this as something like "yeah, it's broken,
but we don't care about this strongly enough to make them fix it."
Remember that in the current text, we're talking about *examples! * of
possibilities to implement the rate-limiter. We'd just be removing
bad examples, and probably adding some non-normative text on why
token-based mechanism is recommended.
This might be a different situation if we said that token-based
rate-limiting MUST be implemented.. but at least I'm not arguing for
going that far.
As for the different values of token-bucket (or timer-based, or
whatever) mechanisms, I'm not sure if that matters that much, it could
be an implementation detail as long as it's reasonable. At most one
could give a "SHOULD be at least XXXX" implementation hint..
--
Pekka Savola "You each name yourselves king, yet the
Netcore Oy kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
