Hello.
Cross posting a find from dd mailing list about bug in delext document.
David Black wrote this clarifications for DELEG:
On 04. 08. 26 16:04, Petr Špaček wrote:
On 24. 07. 26 16:41, Blacka, David wrote:
...
In section 5.1.2, we have:
8. If none of the above applies, SLIST is not modified by this
particular record.
A DELEG-aware resolver MAY implement lazy filling of SLIST, such as
by deferring processing of remaining records, or even individual
names or query types, if SLIST already has what the resolver
considers a sufficiently large pool of addresses to contact.
We think we could use another statement here.
8. If none of the above applies, SLIST is not modified by this
particular record.
Note that the resolution algorithm in [delext section 5.3]
terminates even if the SLIST is empty.
A DELEG-aware resolver MAY implement lazy filling of SLIST, such as
by deferring processing of remaining records, or even individual
names or query types, if SLIST already has what the resolver
considers a sufficiently large pool of addresses to contact.
Note that delext-10 section 5.3 isn't actually correct right now. It
says in step 2.2.3 that if SLIST is empty, remove the leftmost label
and continue. The intent is that if DELEG or NS records are found, the
algorithm stops regardless of the state of SLIST.
These changes were applied to DELEG in
https://github.com/ietf-wg-deleg/draft-ietf-deleg-base/pull/210
Now we need to make sure it is also fixed on delext end.
I've submitted this text change to delext:
https://github.com/RoyArends/draft-ietf-dnsop-delext/pull/2
The proposed text change is this:
-----------
OLD:
2.2.3. If SLIST is not populated, remove the leftmost label from SNAME
and go back to step 2.2, using the newly shortened SNAME. If SLIST is
populated, stop walking up the DNS tree.
-----------
NEW:
2.2.3. If a given SNAME is proven to not have any Delegation Type RRsets
or an NS RRset, remove the leftmost label from SNAME and go back to step
2.2, using the newly shortened SNAME. Note the existence of these RRsets
stops this algorithm even if the SLIST is empty or servers referenced by
SLIST are unusable for any reason.
-----------
That is, is NS exists but is inusable, or DELEG exists and is unusable,
we don't want to walk the DNS tree up in search of 'usable' RRs.
An example of this is 'NS .' or 'DELEG \# 0' which cause SLIST to be
empty but do not permit retry higher up the tree.
--
Petr Špaček
_______________________________________________
DNSOP mailing list -- [email protected]
To unsubscribe send an email to [email protected]
