This is a discussion on Re: Large EDNS0 buffer sizes and IPv6 - DNS ; At Wed, 06 Aug 2008 09:50:30 +0200, Florian Weimer wrote: > > I have no doubt that some vendors lack it (e.g., Microsoft doesn't > > implement it as far as I know), but referring to Linux in this context ...
At Wed, 06 Aug 2008 09:50:30 +0200,
> > I have no doubt that some vendors lack it (e.g., Microsoft doesn't
> > implement it as far as I know), but referring to Linux in this context
> > could actually be misleading; the Linux IPv6 protocol stack (including
> > socket API extensions) had been a cretaceous relic largely behind many
> > other OSes, until the USAGI project incorporated their effort to the
> > mainline kernel.
> Okay. If my stack lacks it, what should I cap the EDNS0 buffer size
> to? 1232 (1280 minues 40 minus 8)? 1200?
I don't think the EDNS0 buffer size matters in this context: it's the
receiver's (=client's) buffer size, while it's the sender that needs
(or has to deal with the case with missing) IPV6_USE_MIN_MTU. Or are
you talking about something like BIND9's specific feature
"max-udp-size"? If so, I'd say it should be not larger than 1232 (if
there's any possibility that some extension headers can be inserted to
DNS responses, it should be smaller accordingly).
Internet Systems Consortium, Inc.
to unsubscribe send a message to email@example.com with
the word 'unsubscribe' in a single line as the message text body.