In your previous mail you wrote:

   On Tue, 6 Jan 2004, Francis Dupont wrote:
   >  In your previous mail you wrote:
   > 
   >    I would like to hear from people who would like to keep the
   >    Timer-based and Bandwidth-based methods with logical reasons.
   >    
   > => Even if Token-based is the best method it can be a bit expensive
   > to implement (code size and memory requirements) on very small devices
   > so I am not convinced we should remove simpler methods... IMHO the
   > Timer-based method with proper parameters (i.e., one ICMP per tick
   > with a 50Hz clock) should be kept. Note that very small devices are
   > not router in general (and never core routers, cf traceroute concerns).
   
   I have to disagree here.  Even if a very small device was not a
   router, it would have to respond to traceroutes etc., even though with
   lower probability that multiple ones would be happening
   simultaneously.  But still, if e.g. a "dumb" host is connected to a
   network, and someone close by (e.g. 3-4 hops) sends a traceroute, the 
   packets could be sent e.g. 1 ms apart.  Any timer-based mechanism is 
   inappropriate, unless it has some kind of "burst allowance", which was 
   what Robert Elz also argued for -- but that's token-based filter just 
   with different words.
   
=> it seems you are confusing ping and traceroute: even a (too) stupid
rate-limiting device is never a problem on the target of a traceroute
because when one'd like to test reachability one uses ping (with the
standard one second delay between probes).

   So, I'd say, keep only token-based filter as an example, but if it
   seems important one could explain, e.g. in an appendix, why
   timer/bandwidth -based mechanisms aren't sufficient on their own.  In
   any case, I believe it's important to document that being able to send
   the packets in burst is an important characteristic of the filter.
   
=> IMHO traceroute is not the good argument here.

   > PS: this topics seems more a v6ops one, doesn't it?
   
   The operational requirements (traceroute, etc.), yes, but this is 
   being revised at this WG so it should probably be OK..
   
=> this is an operational issue so the requirements should come from
the v6ops WG and then we could see if token-based method should be made
mandatory. Because today the only argument we have is for rate-limitation
using any reasonable method.

Regards

[EMAIL PROTECTED]

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[EMAIL PROTECTED]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to