Re: [dnsext] Some feedback on draft-andrews-dnsext-udp-fragmentation-00.txt

[Mark Andrews <[e-mail]>]

In message <[e-mail]>, Fernando Gont writes:
> Mark,
> 
> Here's some feedback about the aforementioned I-D.

Thanks.
 
> Section 3:
> >    It should be noted that even with IPV6_USE_MIN_MTU set to one that a
> >    PTB message may still be received [RFC 2460] which requires a IPv6 to
> >    add a Fragmentation header to subsequent packets.  There is currently
> >    no way to avoid this, without using raw sockets, as there is no way
> >    for a application to request that a Fragmentation header be added to
> >    a packet.
> 
> I understand that this could (at leasst in theory) happen but, out of
> curiosity: are there any devices that leverage the aforementioned
> featuer specified in RFC 2460? NAT 64? Othes?

Its been told to me that there are small (<1280) MTU networks that
don't fragment and reassembly IPv6 at the link layer.  I have no
reason to disbelieve this.  Instead they depend on this.  I have no
direct knowledge.

I do have direct knowledge of DNS/UDP responses being fragmented
at ethernet MTU (by looking at final fragment offsets) rather than
1280 and only the last fragment making it through as a result of a
6in4 link in the path even with repeated queries to the authoritative
server.

Mark
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: [phone]                 INTERNET: [e-mail]