[j-nsp] BGP mtu-discovery and current behaviour

Daniel Roesen dr at cluenet.de
Tue Nov 16 13:55:07 EST 2004

On Tue, Nov 16, 2004 at 10:15:17AM -0800, Pedro Roque Marques wrote:
> > Mmm, yes :-).  But what size does the BGP limit the packet size when
> > it's disabled?  512 bytes as suggested by someone?  The MTU of the
> > egress interface used to send the packet?  Something else?
> Pekka,
> It is the default *BSD TCP behaviour... 512 is the default for
> non-connected links. If directly connected the mtu is used.
> Just what you would expect from a regular BSD stack (or derivative).

Yep, and this is sane because the minimum guaranteed MTU on IPv4
is in the high 500s (IIRC 576), so OS stacks _have_ to stay below
that if PMTUD is not done.

The question is not "why did Juniper use low MTU" but actually "why
didn't Juniper enable PMTUD before". :-)

Best regards,

CLUE-RIPE -- Jabber: dr at cluenet.de -- dr at IRCnet -- PGP: 0xA85C8AA0

More information about the juniper-nsp mailing list