glue record

Mark Andrews marka at isc.org
Fri May 29 23:51:36 UTC 2009


In message <20090529123246.GB8139 at vacation.karoshi.com.>, bmanning at vacation.kar
oshi.com writes:
> On Fri, May 29, 2009 at 02:48:33PM +0700, Anton Zimm wrote:
> > I get this: MOBI servers are not authoritative for push.mobi zone,
> > ns1.push.mobi is authoritative for it.
> > 
> > But since ns1.push.mobi is inside push.mobi zone, this create circular
> > reference. Afaik to solve this circular dependency, there has to be a
> > Glue record for ns1.push.mobi somewhere in root nameservers or mobi
> > nameservers.
> > 
> > I can not find the Glue Record. Where is the Glue Record?
> > 
> > Anton.
> 
> 
> 	what you are looking for no longer exists.
> 	there is no "glue" record per se. As Florian pointed out,
> 	a couple of implementations do what you expect, but BIND 4.x
> 	varients, esp at the TLD level, are scarce.

	BIND 4.x and BIND 8.x were not RFC 1034 compliant in how
	they returned glue records due to a internal design problem.
	This problem was fixed in BIND 9 by seperating the cached
	data from where the zone data was held.
 
> 	As has been explained to you several times, modern thinking
> 	has settled on one (not particularly obvious) way to deal w/
> 	the circular dependence problem.  It really is better to only
> 	have to think about this problem space in one way.

	RFC 1034 says that glue should only be returned in the
	additional section.  There was just a lot of bad implementations.

	Note: RFC 1034 is incorrect when it says that glue is *only*
	required when the namesever's name is below the zone cut.
	That is the case when it is know for certain that glue is
	needed.  There are other delegation patterns that also need
	glue to be returned.

	Mark
 
> --bill
> 
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org




More information about the NANOG mailing list