[00:12:10] cdl leaves the room [00:14:18] cdl joins the room [01:05:04] lengyel leaves the room [01:23:55] Simon leaves the room [01:24:23] Simon joins the room [01:28:04] Simon leaves the room [01:30:46] cdl leaves the room [01:34:18] cdl joins the room [01:37:39] cdl leaves the room [14:51:22] david_partain joins the room [14:51:53] david_partain has set the subject to: IETF73 NETMOD Session 2, Nov 21 0900 - 1130 [14:55:49] mark.ellison@jabber.org joins the room [14:56:28] mark.ellison@jabber.org is now known as Mark Ellison [14:57:27] andy joins the room [15:00:53] lengyel joins the room [15:01:26] yes [15:01:32] sound check? [15:01:46] yes I am [15:02:31] lhotka joins the room [15:02:34] muy beuno senior! [15:04:51] Juergen Schoenwaelder joins the room [15:05:00] good morning [15:05:05] audio is fine here [15:05:06] howdy [15:05:10] excellent [15:05:29] martin: open issues on yang [15:06:51] we're working on mic issues here [15:07:11] slide 2 "Canonical form of data types 1(3)" [15:07:22] ray joins the room [15:07:42] 2(3) [15:08:15] 3(3) [15:13:27] Megumi joins the room [15:13:37] thoughts from radio land? [15:14:22] My preference list is #1, #3, #4, #2 (in decreasing order). [15:14:29] what is the pattern for accepting mixed case DNS names but the canonical form is all lowercase? [15:14:47] And yes, there is an IETF standard that defines the multiple representations of IPv6 addresses [15:15:37] And I believe we do well by supporting the existing standards. [15:17:04] YANG also allows hex numbers. Are these allowed in XML instance documents or just YANG documents? [15:17:22] andy, i'm going to ask your question in just a sec [15:18:39] omplicated [15:18:44] complicated ;-) [15:19:26] OK [15:21:41] last time we agreed that pattern are ANDed - so we change the direction? [15:22:26] bert joins the room [15:22:34] I like Phil's idea and the pattern description clause can cover any cases not obvious [15:22:52] patterns are ANDed [15:23:45] then I prefer (1) [15:27:51] not necessarily the same; the canonical pattern might be much simpler [15:29:35] +1 [15:31:39] keyref vs leafref [15:33:33] dquigley joins the room [15:33:47] dquigley leaves the room [15:36:11] when are must-stmts in non-config nodes evaluated? There is no commit or edit-config to trigger it. [15:39:22] must-stmt on non-config nodes within the target database does not make sense to me; [15:39:35] i support leafrefs [15:39:36] +1 [15:39:59] conformance statement 1(4) [15:41:37] 2(4) [15:42:19] 3(4) [15:43:08] 4(4) [15:43:48] Is there a way that features and capabilities can be specified in a conformance statement? IMO, a complete view of the conformance requirements is needed. [15:44:58] I agree a complete (good) solution can wait until later [15:45:18] less complexity is a good thing; if we do feature, skipping conformance statements is nice [15:47:10] schema discovery [15:47:29] wow... we agreed NOT to include something :-) [15:48:04] no submodules in schema discovery, this is not needed since the include stmts in the main module will identify all the versions of the submodules [15:49:46] andy: comments on martin's answer? [15:51:03] if the submodules can be identified (via tree placement or some flag) then this is OK [15:52:20] how do you know entry foo-sub belongs to foo? Cannot rely on naming conventions [15:56:20] optional belongs-to leaf? [15:57:19] the NETCONF DM should explicitly support submodules if this is important [15:59:50] YANG draft MUST say how to advertise modules and features in the [16:01:41] The schema list might have foo v1.0, foov1.1, etc. -- you have to crack open every single file to see which version imports/includes are needed. [16:05:15] option c: error-path (set to /foo/bb in this example) [16:06:57] andy... i was momentarily out... i'll read what you wrote [16:07:50] error-path instead of bad-element [16:08:21] option B is OK [16:08:41] I prefer to have the errors buffered and returned at the end of the operation, outside the content. [16:12:36] assigned-by [16:13:57] what if there are complex conditions that cause the system to decide to create or not? [16:14:40] is it MUST be filled in, or MAY be filled in by the agent? [16:19:10] this problem goes away when with-defaults is added to the protocol [16:21:40] assigned-by only makes sense if no default clause and mandatory false [16:22:52] +1 [16:26:50] this 'locked knob' feature is used a lot in MIBs; it is up to the DM designers to decide if changing a knob on the fly is too hard or should not be allowed [16:34:58] sorry... didn't see your comment. do you want me to say it even though we're on the next topic [16:34:59] ? [16:35:04] no [16:35:14] sorry [16:35:54] we're on the first actions slides [16:36:01] ^slides^slide [16:36:41] last slide [16:37:12] this is too complicated and can wait until 1.1 [16:39:34] can an action change the target config (candidate or running or startup)? [16:40:57] this needs a new NETCONF capability which should not be defined in the YANG language specification and I am in the camp with those who prefer to do this in YANG 1.1 - our target is to have WG last call in March, no? [16:45:54] Is 'action' allowed to be in a container? nested within lists? What XML is generated then? Does it always start with the parent of the action node? [16:46:24] Exodus joins the room [16:47:36] Exodus leaves the room [16:48:38] right. this is more complicated than it appears [16:48:41] no in 1.0 [16:48:50] +1 (not in 1.0) [16:49:41] typed extension slides [16:49:58] slide 2 [16:50:10] Exodus joins the room [16:51:10] Exodus leaves the room [16:51:40] not needed in 1.0 [16:53:39] what about extensions that are containers and expect further extensions to be present as leafs? [16:54:40] the main usage of Yang is not going to be defining extensions - I fail to see why optimizing this brings us any real benefit [16:57:10] I agree with Juergen, extension args as strings is OK [16:57:22] Megumi leaves the room: Computer went to sleep [16:58:14] The extension statement declares the existance of the extension; it does not try to define it. [16:58:47] +1 [17:03:00] I use the description clause to document these things; that is good enough [17:04:28] They are defined to be not part part of the language; I love extensions as metadata hooks for tools [17:04:59] If you want to buy a car but you can't afford it now, what is the value of buying already some tires? [17:08:21] wait [17:08:23] +1 [17:09:51] which preso? [17:11:37] sorry.. lada's #2 [17:12:02] http://www3.ietf.org/proceedings/08nov/slides/netmod-2.pdf [17:12:42] slide 2 [17:13:39] unreadable is OK for machines [17:15:22] slide 5 [17:15:36] I do not understand what role the 'agent' has in this discussion. The agent implements DMs however it wants and speaks XML instance documents. There is no YANG, XSD, or RNG in the protocol [17:16:47] I agree - the agent validates and how the agent does it is an implementation choice. [17:17:17] i'll read those momentarily... [17:19:27] the agent conforms to the YANG module [17:21:04] how does DSDL help with the not yet present interface? [17:22:42] perhaps I do not understand what the problem is [17:22:55] ok [17:23:03] slide 6 [17:23:47] back to slide 2 [17:25:46] slide 11 [17:28:31] slide 2 [17:28:42] slide 13, i mean... [17:28:49] 14 [17:29:30] we discussed this yesterday - why again? [17:29:55] in case there are new people... i don't know [17:31:11] yang-03 will be stable, yes? [17:31:19] lhotka leaves the room [17:31:22] Mark Ellison leaves the room [17:31:40] andy leaves the room [17:31:45] david_partain leaves the room [17:32:24] ray leaves the room [17:33:08] lengyel leaves the room [17:34:38] Juergen Schoenwaelder leaves the room [17:35:02] bert leaves the room: Computer went to sleep [19:04:13] bert joins the room [19:19:34] bert leaves the room: Replaced by new connection [19:19:35] bert joins the room [19:25:24] bert leaves the room [21:27:24] Exodus joins the room [21:27:30] Exodus leaves the room [21:27:53] mehmet_ersue joins the room [21:28:00] mehmet_ersue leaves the room