Re: [DNSOP] I-D Action: draft-ietf-dnsop-rrserial-00.txt

[Stephane Bortzmeyer <[e-mail]>]

On Mon, Jun 14, 2021 at 10:03:22AM -0400,
 Hugo Salgado <[e-mail]> wrote 
 a message of 55 lines which said:

> In the case of NXDOMAIN, the reason for not adding RRSERIAL is
> because the response already has the SOA in the AUTHORITY, which
> would make it redundant.

OK, I see. Here are two implementations of the client part, in Python
and in Go, using this algorithm (RRSERIAL if NOERROR or SERVFAIL, and
the SOA record if NXDOMAIN).

Python :

% ./test-rrserial.py 200.1.122.30 dateserial.example.com TXT
dateserial.example.com. 43200 IN TXT "Test zone for RRSERIAL record"      
Serial of the answer is 2018081401

% ./test-rrserial.py 200.1.122.30 dateserial.example.com LOC
No value for dateserial.example.com/LOC
Serial found from SOA record: 2018081401

Go :

%   ./test-rrserial 200.1.122.30 incserial.example.com  MX
incserial.example.com.	43200	IN	MX	10 mail.incserial.example.com.
EDNS rrserial found, "1"

%   ./test-rrserial 200.1.122.30 incserial.example.com  SSHFP
Empty answer received
EDNS rrserial found, "1"

Note that you can do it with dig alone but the display of unknown data
is less pretty:

% dig +ednsopt=65024 @200.1.122.30 dateserial.example.com
...
;; OPT PSEUDOSECTION:
; EDNS: version: 0, flags: do; udp: 4096
; OPT=65024: 78 49 7a 79 ("xIzy")

Note also there is a bug in your server: when the name does not exist,
it sends incorrect messages. dig says "Message parser reports
malformed message packet" and the Python library "DNS message is
malformed". And the Go library:

%   ./test-rrserial 200.1.122.30 doesnotexist.incserial.example.com
Error in query: dns: overflowing header size