Minutes of the IETF 85 NETCONF WG Session November 8, 2012 ========================================================== Actors: AB = Andy Bierman AC = Alexander Clemm AL = Alan Luchuk BC = Benoit Claise BW = Bert Wijnen DR = Dan Romascanu JQ = Juergen Quittek JS = Juergen Schoenwaelder KW = Kent Watsen LL = Ladislav Lothka MB = Martin Bjorklund PS = Phil Shafer WS = Wes Hardaker There were approx. 47 participants in the 1,5 hour NETCONF session. Minute takers: Lada Lothka and Juergen Schoenwaelder. * WG Status: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-0.ppt) ME gave the WG status and listed agenda items. The session agenda is available at: http://www.ietf.org/proceedings/85/agenda/agenda-85-netconf * NETCONF over TLS (RFC 5539bis) - Jürgen Schönwälder (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-1.pdf) slide 7: AB: slides talk about rev. -01 but -02 is on the web JS: No, revision of this draft is -01. Robert?: worked on SNMP over DTLS, the extraction of name from the certificate is described there AL: confirming this slide 8: AL: Using preshared keys is optional. We have an optional function to extract name from cert. Preshared key is another opt. capability. The question is whether each server can be expected to be implement these options. 5539bis draft will be updated to address comments and issues discussed. After that there will be another WGLC. * Interop Testing Report: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-3.pdf) slide 7: BW: advancement of RFCs? DR: but now we have two tiers of advancement BW: For some RFCs we can probably claim a large base of customers, but it depends on reports we get, we may even not go for the advancement if the report doesn't provide enough evidence. We want to get public statements on deployment. BW: Who wants to advance these documents? The room was in favour, 8-9 agreed, no objections. DR: Do we need to document wide-spread deployments? BW: Yes, we need to collect information about deployements. Bert will call for confirmation or objections to advance docs on the mailing list. write up a interoperability report. * Andy Bierman presented on Get2: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-2.pdf) LL: Along with limiting depth of the returned tree, it would be useful to be able to limit the number of sibling nodes, either in an analogy to Unix "head" command, or implement some kind of paging. Currently the generic method is problematic for retrieving the contents of a large routing table. AB: Yes, I hear many requests for such a feature. KW: Do the timestamps mean that the server has to remember history of changes? AB: No, there is no remembering of the history, you can only say: "Give me all nodes that changed since ...". KW: We implement systems with one host and multiple logical systems, where the logical systems inherit some properties from the logical system. For this it would be useful to distinguish host system specific timestamps. BW: who wants get2? Several supporters, nobody strongly against it. WH: cannot it be just update to get? BC: get2 solves some but not all issues that are currently on the table. I recommend to try to address all known issues before updating NETCONF. MB: I think we need something like get2. Get2 seems to be addressing a set of problems we want to solve. 15 in favor, no objections. * Operational State Discussion: Juergen Schoenwaelder presented on NC OPS Questions: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-4.pdf) No Q/A. Lada Lhotka presented on NC OPS Use Cases: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-5.pdf) AB: The idea of directly editing state is scary. For example, an operation that clears counters affecting oper. state in a very controlled way, which is very different from writing straight to oper. state. PS: You never clear the counter, you set a new baseline. Phil Shaefer presented on NC OPS - A New Word: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-6.pdf) MB: Lagom tagging is independent of config true/false. ME: I personally would prefer a self-explaining term. LL: This provides the fast path but does it also cover the problem of twin nodes (config & oper. state version)? PS: No. AB: I see a lot of handwaving statememt, a draft requires working out many details. AB: It is unclear how the data is consumed that exists in the Lagom datastore. PS: We only describe how to put data into the Lagom datastore, how the data is consumed is by design implementation specific. AC: How does data in Lagom and data in configuration datastores interact? PS: The desired state gives you insight into what is being finally used. AC: There is an issue of consistence, you may have one thing in running datastore and something different in lagom. Which version take precedence? BC: We should agree what we want to solve before talking about solution. AB: Can I get feedback on whether something like get2 is wanted? I heard only positive feedback on the I-D on the mailing list. BW: Who is interested on the get2 operation? MB: Something like get2 is needed, not sure about all the details. ME: Move to ML, new and possibly joint drafts are appreciated. * Juergen Quittek presented on ONF and OpenFlow config: (http://www.ietf.org/proceedings/85/slides/slides-85-netconf-7.ppt) LL: There was no need to translate YANG to XSD because we have RFC 6110. JQ: 6110 implementations are scarce and are hard to use. LL: People likely specifically wanted XSD, there is a standards-track translation to RelaxNG and Schematron. JS: Beep to historic is a done deal. With TLS using certificates on both endpoints, doing call home might not be a big deal. KW: I wrote a draft on Reverse SSH which supports call-home but is now expired. ONF have a 'call home' use case and might ask the Netconf WG to find a new way to support it after declaring BEEP as 'historic'.