On Wed, Nov 10, 2010 at 01:36:32PM +0000, Dave Thaler wrote:
> It is my understanding that if the string passed to a name resolver library
> contains dots, no search list should be used. (Reference: RFC 1535)
And then there is the unix world of stub resolvers, where the search
string is used after an initial attempt on any query with one or more dots
("ndots:1" is the default configuration).
And yet it is unlikely that someone is going to name their AFTR endpoint
'example.com'.
Consequences;
1) The root receives queries for (and formulates NXDOMAIN responses)
'aftr.site'. The client then moves on to try its search path. Despite
negative cache entries in recursive resolvers, this is littering.
2) Windows (no search behavior on any dots) and Unix (behavior I
described) becomes incompatible. The two types of client, given
the same aftr-name option contents will process these contents in
different ways, either requiring specialized configuration, or
defining a standard behavior by fiat or muscle.
So I believe a standard description of the process that both can implement
equally (with the same consequent behaviors) and does not trouble the root
nameservers needlessly is in order.
I think what is essentially a "ndots:2" configuration setting, which I
have verbosely described, is the correct method.
--
David W. Hankins "If you don't do it right the first time,
Network Mercenary you'll just have to do it again."
-- Jack T. Hankins
_______________________________________________
Softwires mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/softwires
