e07@nikhefh.nikhef.nl (Eric Wassenaar) (04/19/91)
Suppose this is the zone data file at 'host1.foo.bar' $ORIGIN foo.bar. @ IN SOA ... IN NS host1.foo.bar. host1 IN A .... host2 IN A .... sub2 IN NS host2.foo.bar. sub1 IN NS host3.sub2.foo.bar. The question is here whether or not to add a glue record for 'host3.sub2.foo.bar' on behalf of subsequent queries such as 'host4.sub1.foo.bar'. In principle, after a few iterations, the nameserver would be able to figure out the address of host3 himself, so that a glue record would not be absolutely necessary. But will he actually carry out these iterations ?? If he does, adding the glue record would save such steps. Is a glue record for host3 in this case a. absolutely necessary b. redundant but a time-saving handy shortcut c. plainly wrong Eric Wassenaar -- Organization: NIKHEF-H, National Institute for Nuclear and High-Energy Physics Address: Kruislaan 409, P.O. Box 41882, 1009 DB Amsterdam, the Netherlands Phone: +31 20 592 0412, Home: +31 20 6909449, Telefax: +31 20 592 5155 Internet: e07@nikhef.nl
rickert@mp.cs.niu.edu (Neil Rickert) (04/20/91)
In article <1210@nikhefh.nikhef.nl> e07@nikhefh.nikhef.nl (Eric Wassenaar) writes: >$ORIGIN foo.bar. >@ IN SOA ... > IN NS host1.foo.bar. >sub2 IN NS host2.foo.bar. >sub1 IN NS host3.sub2.foo.bar. >The question is here whether or not to add a glue record for >'host3.sub2.foo.bar' on behalf of subsequent queries such as >'host4.sub1.foo.bar'. >Is a glue record for host3 in this case >a. absolutely necessary >b. redundant but a time-saving handy shortcut >c. plainly wrong My vote is for (c.). Otherwise when the data (for example the A-record) for host3 is changed, somebody has to remember to update the glue record, and it is too easy for that to be overlooked. In most practical situations the current host would probably be a secondary nameserver for the sub2 and sub1 subdomains, in which case it would happen to have the necessary A-record available anyway. -- =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= Neil W. Rickert, Computer Science <rickert@cs.niu.edu> Northern Illinois Univ. DeKalb, IL 60115 +1-815-753-6940
sra@LCS.MIT.EDU (Rob Austein) (04/22/91)
The rule for glue RRs is that you should only add them if there is no other way that resolvers will be able to contact the nameserver to which you delegating. Since, as you say, a resolver could find the address of the delegated nameserver in your example, there is no need for a glue RR. That being the case, you should not add a glue RR. See RFC-1033 for a more complete discussion of glue RRs. --Rob Austein