NSEC3 Update IETF 66, Montréal David Blacka, Verisign Review • NSEC3 is an NSEC alternative that: – Prevents zone enumeration by using hashes of domain names – Optional optimization for delegation centric zones called Opt-Out. What is going on? • • • • Workshop in Frankfurt on May 8-10 Draft is now at version -06 Side meeting this Monday Proposed next workshop Frankfurt Workshop • • • • • • Purpose: interoperability testing, discussion. Testing signing, serving, validation. No major issues found. But didn’t test everything. Created a semi-permanent test environment Notes for the workshop: http://www.nsec3.org/cgi-bin/trac.cgi/wiki/Workshop1Notes Frankfurt Workshop, cont. • Tests that still need to be done: – NSEC to NSEC3 rollover, vice versa – Signaling mechanism(s) – Traversing down various combinations of NSEC, NSEC3, with/without Opt-Out – Spoof tests, esp. Wildcards + Opt-Out. Changes from -05 to -06 • • • • • Lots of wordsmithing. Added hash length field to RDATA. New section on signaling. New section on dynamic update. New text on handling unknown NSEC3 hash algorithms • Updated examples. Changes, cont. • Responses are now required to use NSEC3 with all the same parameters. • A few things still missing: – Text on transitioning a zone from NSEC to NSEC3. – Open issues. Issues • NSEC3 has an issue tracker – http://www.nsec3.org • Some issues are closed. – This means that the draft editors think that the issue is addressed. – Not that the issue cannot be discussed further. Open Issues • Issue 8: Signaling – This is about interoperability with RFC 4035-based resolvers. I.e., NSEC3 signed zones should be treated as insecure. – At workshop, discussed two possibilities – New draft describes the use of unknown signing algorithms. – Not set in stone, but that is what has been implemented. Open Issues, cont. • Issue 9: Iterations upper bound – Document sets an upper bound based on RSA signing times, resolvers may treat NSEC3 RRs with too many iterations as BOGUS – Should be based on verifications instead? – Resolvers should treat as INSECURE instead? – How does upper bound change over time? Open Issues, cont. • Issue 11: Queries for NSEC3 ownernames – 3 different approaches have been suggested. – All 3 have been described in past versions of the draft. – Main issue is if direct queries for NSEC3 RRs should work. Open Issues, cont. • Issue 18: Signaling complete NSEC3 chains. – So auth servers (primary and secondary) can determine the NSEC3 parameters – Currently requires finding the NSEC3 for the zone apex. – 3 other proposals: new zone apex RR (NSEC3PARAM), reuse NSEC3 at zone apex, special case the zone apex NSEC3. – Currently heavily leaning toward using NSEC3PARAM Open Issues, cont. • Issue 22: Separating NSEC3 algorithms from DS algorithms – Currently re-uses the DS hash algorithm registry. – But, desired hash properties are not (exactly) the same. – Do not necessarily want to define new NSEC3 hash when DS defines a new hash. – Use of some hashes might require additional specification. Fin Questions/Comments?