Hi, while debugging the Google/IPv6 issue, we discovered something strange. Our uplink provider operates caching DNS servers, and they reply with a rather detailed Additional section when asked for MX records, but only with cached results.
For example, if example.com has an MX record pointing to mx.example.com, and mx.example.com has one A and one AAAA record, then the caching DNS server will return as many of those records as it has cached in memory. As most systems using the cache seem to only ask for A records, A records appear to be cached more often than AAAA records, but that is irrelevant. Most tools, mainly libc's resolver, seem to ignore the Additional section and resolve relevant names on their owns, explicitly asking for the RR types they are itnerested in, and that's what seems to be appropriate. Postfix, however, seems to rely on the Additional section (if it has at least one RR for the MX host?), missing out on any records that might be there but not cached by the uplink DNS server. We do not quite see an situation where this might break badly, because normally one MX result is to be considered as good as any other, but I still wanted to ask whether this behaviour is intentional and the limitations are known. Cheers, Nik -- <Natureshadow> Auf welchem Server liegt das denn jetzt…? <mirabilos> Wenn es nicht übers Netz kommt bei Hetzner, wenn es nicht gelesen wird bei STRATO, wenn es klappt bei manitu. PGP-Fingerprint: 3C9D 54A4 7575 C026 FB17 FD26 B79A 3C16 A0C4 F296
signature.asc
Description: Digital signature