MSIG proposal (on-the-fly sigs for ordinary records) Was: DNSSEC is hard to get right

[Stephane Bortzmeyer <[e-mail]>]

On Tue, Aug 31, 2010 at 02:55:08PM +0800,
 Jiankang YAO <[e-mail]> wrote 
 a message of 11 lines which said:

> I propose a lightweight DNSSEC.
> 
> http://www.ietf.org/id/draft-yao-dnsext-msig-00.txt

I've just read the draft and I'm not sure of the problem it intends to
solve. There are two parts where DNSSEC could be regarded as "too
heavy":

1) Administrative procedures, key management, resigning, etc.

2) Work for the name servers (loading large zones, sending large
packets, validating, etc).

MSIG addresses only the second. The first one, which was the cause of
the failure for iab.org, is exactly the same as with the current
DNSSEC.

Even for the second, MSIG addresses a problem that we do not feel (for
the signing of .FR, which will be on line next week, the size of the
zone was the smallest problem) and creates a new problem: the
authoritative name server now must generate a signature for every
request! You will eat less RAM but use much more CPU.

Also, if I understood the draft correctly:

* Every authoritative name server, even a slave, will require a copy
of the private key (since it will have to sign the responses
on-the-fly). Bad for manageability and security.

* MSIG secures the link from the authoritative name server to the
resolver but cannot help if there are chained resolvers, or cannot be
used for the last mile. (I'm not sure about this last point, it is not
clear in the draft.)