[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: comments on draft-daniel-ipv6-ra-mo-flags-00.txt



Hi Fred,

Fred Templin wrote:
> Greg Daley wrote:
> 
>> For MTU, it's clear that you need to take the smallest
>> (most restrictive) value advertised.  This is because
>> choice of a higher MTU is likely to have worse effects
>> than 
> 
> 
> 
> I think this needs a bit of refinement. For multicast RAs (both unsolicited
> and solicited), the MTU option gives an indication of how much the router
> itself can accept - presumably as a reflection of the router's attached 
> link
> MTU, router buffer restrictions, etc. For unicast, solicited RAs, however,
> the MTU option gives an indication of how much the forward path from
> the soliciter to the router can accept.
> 
> Since the L2 switches/bridges/etc. on the paths between a soliciter and
> its candidate routers may be quite different, it seems reasonable that
> the soliciter might want to store different MTUs reported by different
> routers in solicited unicast RAs.

Indeed, when you're commuicating off-link, this may be correct.
When communicating within the link (link-local or to neighbours)
you can't be guaranteed where the packets will go within the domain.

The smallest option is the safest in this case,
since no specific information about the path taken is available.

Path MTU is always based on the narrowest MTU on a path.
Within a link, the same applies (narrowest MTU on path within the link),
regardless of what the attached interface on a router believes.


Greg


--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------