DNS over TCP (was Re: Offence CNAME as MX??)

Jim Reid jim at rfc1035.com
Wed Dec 19 14:43:24 UTC 2001


>>>>> "Simon" == Simon Waters <Simon at wretched.demon.co.uk> writes:

    Simon> Still part of the reason for following standards is not
    Simon> only to ensure that archaic versions of sendmail can e-mail
    Simon> you, but that future modifications to the protocol can be
    Simon> cleanly applied. Witness the importance of accepting TCP
    Simon> queries to handling DNSSEC and other DNS protocol
    Simon> extensions sensibly.

Er, not exactly. DNSSEC undoubtedly makes DNS packets bigger. But that
does not necessarily mean DNSSEC-aware traffic has to use TCP. It can
go over UDP when EDNS0 is used to arrange DNS datagram payloads bigger
than 512 bytes. And could you tell us what other DNS protocol
extensions (plural) are expected to make TCP queries more likely or
necessary?

I agree with your argument for following standards. ie To make future
protocol extensions simple, but not to ensure interoperability with
ancient and probably buggy versions of sendmail. Or other legacy code.


More information about the bind-users mailing list