<<< Chronological >>> Author Index    Subject Index <<< Threads >>>

Re: rfc1916 or rfc2916bis and DNS packet sizes

On 31 Mar 2004, at 11:01 am, Jim Reid wrote:

"lwc" == Conroy, Lawrence (SMTP) lwc@localhost writes:
    lwc> Thus it's only "easy" for those folks having a single contact
    lwc> - remember that some of the old clients only use UDP queries
    lwc> (without EDNS0 support).

A host that cannot do DNS lookups over TCP is broken and needs to be
fixed. If the query response is bigger than the standard maximum
payload of 512 bytes, the server sends a truncated response. The
client is then supposed to repeat the query over TCP to get all the
data. In this context EDNS0 is just a way of negotiating a bigger UDP
payload. EDNS0 can't guarantee that truncated responses won't ever
happen, though it should make them less likely. Therefore support for
queries over TCP is still needed.

Truncated responses wouldn't be too hard to achieve in ENUM even with
an 8K EDNS0 UDP packet. Mix a delegation holding a handful of NAPTRs
with DNSSEC and you're just about there.

Hi Folks,
  one word (well, acronym): GPRS
  another one: JAVA
In an ideal world, all things are possible. In a cheap handy, however...

(in an ideal world, DNSSEC would be feasible :).

atb,  L

<<< Chronological >>> Author    Subject <<< Threads >>>