Bind 9, Wildcard Records and Road Runner

Jonathan de Boyne Pollard J.deBoynePollard at Tesco.NET
Fri Nov 12 04:44:06 UTC 2004

MF> If the wcard draft is standardized, records like
MF>       www.*.mp. IN A
MF> become acceptable, true?

False.  They don't _become_ acceptable.  The _already are_ acceptable.

They just don't do what you, no doubt conditioned by all of those
softwares that use wildmat(), think that they do.  But then this is true
of wildcards (in DNS) in general.  They don't do what people expect them
to do.  *Every* DNS server software deviates from the RFC 1034 algorithm
in one respect or other when it comes to wildcards.  And they all do so
in the name of doing what the users actually expect wildcards to do.
(One DNS server software, Microsoft's, even does what people expect,
rather than what the RFC 1034 algorithm dictates, with respect to "MX"
resource record sets and wildcards.)

The irony, of course, is that there is no universal agreement amongst
DNS server softwares on how wildcards *should* work.  They each deviate
from the RFC 1034 algorithm in different ways.  (Microsoft's DNS server
and MaraDNS both also look for wildcards when they find empty resource
record sets.  ISC's BIND processes alias chains when it encounters
wildcard "CNAME" resource records.  Dan Bernstein's "djbdns" allows
wildcards to match leaf nodes even when interior nodes for enclosing
superdomains exist.)

The wildcard clarification is not yet ready for standardization,
incidentally.  It currently contains a few errors.

More information about the bind-users mailing list