Re: CNAME in SRV record RDATA (Target field)

I've tested this on BIND 9.3.1, also on SuSE 9.1, still no errors or
warnings; querying for that SRV record doesn't yield any errors, only
there are no IP addresses in the additional section:

dig @localhost srv _http._tcp.varza

; <<>> DiG 9.2.3 <<>> @localhost srv _http._tcp.varza
;; global options: printcmd
;; Got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 58079
;; flags: qr aa rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 1, ADDITIONAL: 0

;; QUESTION SECTION:
;_http._tcp.varza. IN SRV

;; ANSWER SECTION:
_http._tcp.varza. 604800 IN SRV 10 0 80 www.varza.

;; AUTHORITY SECTION:
varza. 604800 IN NS localhost.varza.

So, is this a good candidate for bind9-bugs or is there an explanation
for this behaviour?

On 18 Apr 2005 06:50:04 -0700, Stefan Puiu <stefan.puiu@gmail.com> wrote:
> Hello,
>=20
> according to RFC2782 (section "The format of the SRV RR", on page 4),
> the last element in the SRV RDATA section, target, is:
>=20
> The domain name of the target host. *There MUST be one or more
> address records for this name, the name MUST NOT be an alias (in
> the sense of RFC 1034 or RFC 2181).*
>=20
> So, as far as I understand, target can't be a CNAME - it can't be an
> alias and it must own at least one address record.
>=20
> However, I created a test zone in BIND including the following lines:
>=20
> $TTL 1W
> @ IN SOA localhost root (
> 3 ; serial
> 2D ; refresh
> 4H ; retry
> 6W ; expiry
> 1W ) ; minimum
>=20
> IN NS localhost
> IN MX 10 mail
>=20
> _http._tcp IN SRV 10 0 80 www
> ktulu IN A 10.0.0.4
> www IN CNAME ktulu
>=20
> When feeding this zone to BIND 9.2.3 (running on SuSE Linux 9.1),
> named starts without reporting any errors or warnings (from
> /var/log/messages):
>=20
> Apr 18 16:11:57 tuxws19 named[16569]: loading configuration from
> '/etc/named.conf'
> Apr 18 16:11:57 tuxws19 named[16569]: zone varza/IN: loaded serial 3
>=20
> Is this behaviour intentionally inconsistent with the RFC (it sounds
> convenient to be able to use CNAMES in the target field, although this
> would make resolution harder for clients) or is it an implementation
> error? I've seen an older discussion on this on the list, with various
> people pointing out that such a setup would be erroneous:
> http://marc.theaimsgroup.com/?l=3Dbi...31813267&w=3D2.
> However, no mention on how BIND handles that. Since there is no
> mention of CNAMES and SRVs in the BIND 9.3.1 CHANGES file (in bugs
> fixed since 9.2.3 came out), I thought this to be legitimate question.
>=20
>