Andy Bierman <andy@xxxxxxxxxxxxx> wrote: > Hi, > > On Thu, Jan 9, 2014 at 7:47 AM, SM <sm@xxxxxxxxxxxx> wrote: > > > Hi Juergen, > > At 02:58 09-01-2014, Juergen Schoenwaelder wrote: > > > >> People want to be able to configure timezones using the timezone > >> naming scheme commonly supported by operating systems (e.g., > >> Europe/Paris). This document defines a serialization of the timezone > >> names into YANG format for this purpose. I believe there was indeed > >> strong concensus on this in the WG. > >> > > > > There was a message about a quick poll [1] to adopt > > draft-lange-netmod-iana-timezones-01 on 2 July, 2012. Three persons > > expressed their support and the WG Chair expressed his support as a > > technical contributor [2]; there weren't any message objecting to adoption. > > There were two comments [3][4] on 30 July, 2012. There was a WG Last Call > > announcement [5] and two reminders [6][7] about it. There was a message [8] > > on 12 November, 2013 from a WG participant (I am not taking into account > > the message from the author and one of the WG Chairs). My comment about > > the determination of consensus being questionable is based on the messages > > in the NETMOD mailing list archive. > > > > This document aims at establishing an IANA maintained serialization of > >> whatever list of names the timezone database contains. It does not > >> change the way the TZ coordinator, an IANA Designated Expert, takes > >> decisions. Perhaps this needs to be stated more explicit. > >> > > > > There is a thread at http://mm.icann.org/pipermail/tz/2012-May/017711.html > > The problem is not changing the way things are done. I would describe the > > current TZ approach as keeping the TZ information up-to-date by posting a > > (software) release every now and then. It works for me. > > > > So far (in the MIB world), the initial versions of IANA maintained > >> modules were published as Proposed Standards on the standards track > >> (although the only thing that really remains valid over time are the > >> IANA considerations, which however often require the initial module to > >> be available). We simply followed this practice. > >> > > > > If I understood correctly the YANG module defined in the draft creates > > IANA-registered timezones based on public domain [9] information about > > timezones. IANA is then asked to keep the IANA-registered timezones > > up-to-date. That sounds okay if the politics about time are not taken into > > consideration [10][11][12]. Yes. The idea is to extract the names from the published TZ database and list them in a YANG module. Nothing more. Entries cannot be added directly to the YANG module; they will always just mirror what's in the TZ database. > > There was concensus in the WG that it is desirable to configure > >> timezones by name (e.g., Europe/Paris) instead of having only hard > >> coded UTC offsets - the benefits should be obvious. > >> > > > > I agree that there can be benefits to that approach. I could not find the > > (mailing list) message where that WG decision was taken. > > > > > After reading the IANA Considerations section closer, I have some concerns > about the use of a YANG enumeration. This section does not say who must > update the enumeration typedef and republish the RFC containing the new > YANG module, every time the TZ database is changed. > Is it the NETMOD WG? IANA? I think it is clear that the intention is that is IANA. But of course it could be clarified. Maybe it should even say that it is the responsibility of the TZ Coordinator. However, this must presumably be verified with him first. > This may not be a practical long-term solution. > The purpose of a YANG enumeration type is to allow a well-constrained > value set that can be machine-validated. If the data type was a string > instead of an enumeration, there would be no IANA coupling or republishing > (or YANG-based machine validation). The description-stmt can say a leaf > using this data type MUST match a value in the TZ database. If it is not feasible to maintain a YANG "mirror" of the timezone names from the TZ database, then I guess we can use a string. I prefer the formal enumeration though. > The draft (sec. 3.1) also says to change the status of the old enum to > obsolete > and add a new enum, in order to change the name. YANG definitions are > supposed to transition from current to deprecated, not current to obsolete. RFC 6020, section 10, explicitly allows a definition to go from current to obsolete: o A "status" statement may be added, or changed from "current" to "deprecated" or "obsolete", or from "deprecated" to "obsolete". /martin