[13:22:22] --- dperkins3600 has joined
[14:09:07] --- sharonchisholm has joined
[14:09:22] <sharonchisholm> Dave is going over the intro slides
[14:09:25] <dperkins3600> hi - hows the weather
[14:09:45] <sharonchisholm> It rained last night, but today it seems nice. Not as humid as I would have thought
[14:10:46] --- mark.ellison has joined
[14:11:55] <dperkins3600> hey - it's not the base items, the issue is the grouping and indexing
[14:12:20] <sharonchisholm> mainly indexing
[14:12:37] <sharonchisholm> yeah, and the same stuff was never related together
[14:13:32] <dperkins3600> As well as getting the contents of a table that has lots of instances
[14:14:07] <dperkins3600> what page is he on now?
[14:14:55] <dperkins3600> got it - thanks
[14:15:45] <sharonchisholm> would you translate snmp types into a readable string or would they still be yucky encoded, like as an octet string?
[14:18:04] <dperkins3600> I hate the performance hit with text based encodings, but once you have signed up for that, you should just encode numbers in text format. for binary items (like octet strings), I'm not sure the best encoding. What do you suggest?
[14:18:48] <sharonchisholm> I think everything should be strings (real types, but it all comes down to being strings in the end)
[14:19:04] <sharonchisholm> <ip></ip>
[14:19:27] --- dromasca has joined
[14:20:08] <dperkins3600> how do you encode a blob of binary info, such as 10 octets?
[14:20:48] <sharonchisholm> you shouldn't send stuf like that, but if you do, make it clear
[14:20:49] <dperkins3600> hey - questions later
[14:20:50] --- atarashi has joined
[14:21:39] --- jeremiah has joined
[14:21:45] <dperkins3600> sorry dan - sharon and I were just chatting
[14:23:39] <dromasca> np, let me know when you have real questions :-)
[14:25:36] <dperkins3600> how many people are at the BOF?
[14:26:03] <jeremiah> probably about 25
[14:27:41] --- bwijnen has joined
[14:29:26] <sharonchisholm> Dan's document was started as a set of datatypes for use in Netconf, not as anything to do with SNMP & SMI
[14:29:37] <dperkins3600> hey - doesn't Mike Kirkum's (?sp) tools do the same MIB module to xsd conversion as smidump?
[14:37:38] <dperkins3600> real question/observation for latter - shouldn't the data type mapping also include considerations for data types found in database systems
[14:37:49] <sharonchisholm> why?
[14:38:40] --- bwijnen has left
[14:39:00] --- dlpartain has joined
[14:40:24] <dperkins3600> Much data is gathered via SNMP, stored in a data base, and periodically analyzed. Having a standard mapping format makes it easier for the exchange of info that has been stored in a data base.
[14:44:33] <sharonchisholm> most systems have a datastore which may not not be a database. This is not exposed to the wire
[14:44:37] <sharonchisholm> or should not be
[14:46:59] <sharonchisholm> slide 2
[14:47:12] <sharonchisholm> 4
[14:47:25] --- dlpartain has left
[14:48:28] <sharonchisholm> 5
[14:50:55] --- bert has joined
[14:51:04] <sharonchisholm> 6
[14:51:12] <sharonchisholm> 8
[14:51:15] <sharonchisholm> 7
[14:51:18] <sharonchisholm> :-)
[14:51:56] <sharonchisholm> Is he presenting on Dan's draft or the general translation one?
[14:55:41] <sharonchisholm> ... the other one apparently.
[14:56:45] --- bert has left
[14:56:52] --- bert has joined
[14:56:56] --- sharonchisholm has left: Replaced by new connection
[14:59:45] --- sharonchisholm has joined
[14:59:46] <dromasca> the other one
[15:00:03] <sharonchisholm> 15
[15:00:27] <sharonchisholm> 16
[15:02:44] <sharonchisholm> 17
[15:03:51] <sharonchisholm> 18
[15:04:35] <sharonchisholm> 20
[15:08:49] <sharonchisholm> 21
[15:11:45] <sharonchisholm> Dave, if you'd like the authentic experience of being in the room turn your AC down to 10 celsius
[15:11:47] <dperkins3600> I missed the end of that, since I got an interrruption. Is it too late for questions?
[15:11:59] <sharonchisholm> on Yan Li, yes
[15:12:10] <bert> http://www3.ietf.org/proceedings/07jul/slides/xsdmi-0.txt
[15:12:12] <dperkins3600> What is DBH doing now
[15:12:15] <bert> that is the charter
[15:12:27] <bert> draft charter I should say
[15:12:34] <sharonchisholm> thanks bert
[15:17:40] <dperkins3600> The issue of "naming" is the really hard part. Row pointer is
[15:18:07] <sharonchisholm> you could translate the row pointer into a keyref
[15:18:15] <sharonchisholm> except there are some scope issues with that
[15:18:33] <sharonchisholm> perhaps the xpath without using the keyref construct
[15:19:41] <dperkins3600> not the only issue, since you don't have to use a row pointer to identify instances. In general you need context and instance info. (yes, these are "key refs")
[15:21:12] <dperkins3600> In Yan's example, how instancing was done was very SNMP specific, and very clumsy. Don't know of the "best" way for identying an instance.
[15:22:31] <dperkins3600> Don't know of the "best" way for identifying an instance in a "protocol independent" way.
[15:24:35] <sharonchisholm> do you want anything asked at the mike?
[15:26:11] <dperkins3600> I'm trying how to figure out a question to ask about naming, and how SNMP references to instances are translated. (Andy's sort of asking the same question)
[15:27:33] <dperkins3600> Are there any concerns based on existing SNMP implementations, or is this work based on new implemenation?
[15:28:12] <sharonchisholm> what do you mean by concerns and do you want me to ask that?
[15:30:03] <dperkins3600> Well, you might make different choices if you already have devices with SNMP support, versus writing code for devices that don't already have an SNMPagent.
[15:30:28] <sharonchisholm> yeah.
[15:30:34] <sharonchisholm> did you want me to ask?
[15:30:56] <dperkins3600> no - don't have a clear enough question
[15:31:03] <sharonchisholm> fair enough.
[15:37:29] <sharonchisholm> Dave - want to contribute?
[15:37:38] <dperkins3600> real comment - again would like to have data types found in data base systems considered in the data type conversion
[15:39:18] <dperkins3600> real comment - The naming (instance specification) and the references to instances seem to be the REALLY hard part. What do other think?
[15:40:40] <sharonchisholm> lots of hands
[15:40:56] <sharonchisholm> 13
[15:41:10] <dperkins3600> data type work - yes
[15:41:17] --- m_ersue has joined
[15:42:08] <dperkins3600> Just don't believe that data types for references is easy. Maybe need to defer.
[15:42:52] <dperkins3600> What about rowpointer?
[15:44:10] <sharonchisholm> details to be worked out ...
[15:44:12] <dperkins3600> Algo-translation only works for use in an SNMP-based XML protocol - consider references.
[15:44:44] <dperkins3600> Also consider RowStatus, and objects to support actions, like the ping MIB
[15:44:59] <sharonchisholm> 9
[15:45:31] <dperkins3600> no - not possible except for SNMP proto
[15:45:36] <sharonchisholm> pro 3
[15:45:44] <dperkins3600> NOT
[15:45:51] <sharonchisholm> con 4
[15:46:00] <sharonchisholm> (there are 31 in the room)
[15:48:50] <dperkins3600> I can review - but since I'm not an XSD expert, can't write
[15:54:04] --- m_ersue has left
[15:54:23] <sharonchisholm> ok. See you in isms?
[15:55:15] <dperkins3600> Will try, but got to get lunch.
[15:55:41] --- mark.ellison has left
[15:55:45] <sharonchisholm> alright. If not, take care
[15:56:03] --- dperkins3600 has left
[15:56:06] --- sharonchisholm has left
[15:57:14] --- jeremiah has left
[15:59:13] --- bert has left
[16:08:21] --- dromasca has left
[16:11:25] --- atarashi has left
[16:16:12] --- bert has joined
[16:17:05] --- bert has left
[16:57:00] --- dbh2 has joined
[16:57:11] --- dbh2 has left