Test your connectivity for World IPv6 Day

Mark Smith nanog at 85d5b20a518b8f6864949bd940457dc124746ddc.nosense.org
Tue Jun 7 19:50:59 CEST 2011


On Tue, 7 Jun 2011 17:31:11 +0200
Rémi Després <remi.despres at free.fr> wrote:

> 
> Le 7 juin 2011 à 14:29, Mark Smith a écrit :
> 
> > On Tue, 7 Jun 2011 12:30:08 +0200
> > Rémi Després <remi.despres at free.fr> wrote:
> > 
> >> 
> >> Le 6 juin 2011 à 21:41, Daniel Roesen a écrit :
> >> 
> >>> On Mon, Jun 06, 2011 at 05:58:43PM +0200, Rémi Després wrote:
> >>>> Unless there are good reasons to know that a longer PMTU applies to all
> >>>> their connections, all servers SHOULD send IPv6 packets to off-link
> >>>> destinations with 1280 octets as default PMTU.
> 
> I agree that this sentence I wrote is too restrictive.
> Longer PMTU's can be discovered on a per connection basis.
> Yet 1280 remains AFAIK a good value to start with.
> 
> 
> >>> 
> >>> So you advocate making 1280 effectively the maximum MTU, not the minimum?
> >> 
> >> The "default" MTU, not the "maximum".
> >> 
> >> If some PMTUD permits to detect that a larger PMTU is possible on a specific path, it can then be used.
> > 
> > Conventional PMTUD only works downwards, so if you start at a
> > default MTU 1280, you'll never discover a larger PMTU than 1280. MSS
> > also works to constrain that, as it tells the other end the maximum
> > segment size to send. 
> 
> This being so, which size do you recommend to start with in IPv6?
> - 1500?
> - 9000 to get a chance to reach jumbo-frame sizes?
> 

Host's outbound interface MTU. Not so much a recommendation though, as
this is IPv4 and IPv6's PMTUD default behaviour. 

> RD
> 
>  
> 
> 
> > 
> >> Also, on-link connections can use the link MTU.
> >> Regards,
> >> RD
> >> 
> >>> 
> >>> I'm with Fred here. Fix broken paths, not work around the problems.
> >>> 
> >>> Best regards,
> >>> Daniel
> >>> 
> >>> -- 
> >>> CLUE-RIPE -- Jabber: dr at cluenet.de -- dr at IRCnet -- PGP: 0xA85C8AA0
> >> 
> 


More information about the ipv6-ops mailing list