LMAP WG IETF 90,
Toronto, Canada
Thursday July 24, 2014, 0900-1130, Morning Session I
Audio recording at: http://www.ietf.org/audio/ietf90/ietf90-territories-20140724-0900-am1.mp3
Thanks to Barbara
Stark and Phil Eardley for taking notes during the meeting.
Time markings of
the mp3 recording are noted for the start of each topic.
Full names are listed first time a person appears in these minutes. Subsequent
mentions only use first name. There were no duplicate first names among
speakers.
Chairs: Dan Romascanu, Jason
Weil
Minutes
Chairs Slides (at 00:50 in
mp3 recording) [http://tools.ietf.org/agenda/90/slides/slides-90-lmap-3.pdf]
Dan: Welcome. This is LMAP.
Note Well <slide 2>. This is the agenda (slide 3). Agenda bashing? <no
comments voiced>. We have a jabber scribe. Need note-takers. Phil and
Barbara are minute-takers.
[Agenda
from Slide 3, with some names changed to actual presenter:
1. Note Well, Note Takers,
Jabber Scribes, Agenda Bashing - Chairs (5 min)
2. WG Status and Timeline
Update - Chairs (5 min)
3. LMAP Framework - Phil (45
min)
4. LMAP Information Model - Jrgen (20 min)
5. Protocol Contributions
5.1 - HTTP-based protocol -
Marcelo (20 min)
5.2 - HTTP-style control protocol
- Lingli Deng (20 min)
6. Liaison from BBF and
proposal for interim meeting in Dublin - chairs (20 min)
7. Next steps and open mic]
<slide
4> Use cases submitted to IESG. Benoit has
reviewed, comments later today
Framework document completed 3rd WGLC. Info model is progressing. 2 protocol submissions received. Others are still welcome to submit protocols.
Framework - Further Reviews by 31st july. Authors to resolve comments with commenters. Chairs will then decide whether another wglc is needed/
<slide 5> Charter milestones have been
moved out 7 months.
Next on the agenda is framework.
------
LMAP Framework (at 8:20 in
mp3 recording) [http://tools.ietf.org/agenda/90/slides/slides-90-lmap-0.pdf]
Phil Eardley: <slide
2> Many revs, but progress has been made with each.
<walks through slide 3, starting with suppression>
Dan (at mic, chair hat off):
If there is a first suppress message that just suppresses task A and a 2nd
that just suppresses task B, what does this do?
Phil: The 2nd
message will cause A to be re-activated and B to be suppressed.
Dan: This needs to be made
clear.
Phil: OK. And are people
happy with that approach? <There was no dissent.>
Phil: <continues with
slide 3> Are people happy with how suppress now is? <No dissent.>
<describes suppression flag behavior> Happy? <No dissent> <continues
with slide 3 on measurement method>
Greg Mirsky: How is
Measurement Method Role reflected in registry?
Phil: Registry is about
metrics and methods for measuring a metric. Registry doesnÕt mention roles.
ItÕs just a field (in the info model). Registry has one entry for the metric and
if you have separate role in registry it might end up having 2 entries in the
registry?
Marcelo Bagnulo: WeÕll have
2 entries in the registry?
Phil: No, weÕre going to
have one entry in the registry.
Barbara Stark: Then how do
we map from this field, unambiguously to roles defined by the Measurement
Method?
Al Morton: Maybe we can have
column in the registry that mentions the roles that exist for the Measurement
Method? Description of how the role behaves is in Measurement Method.
Mike Bugenhagen: Certain
words might cause confusion if used as role. There is ambiguity if a ambiguous
words are used.
Marcelo: ThatÕs in
description of Measurement Method
Mike: Concern is language in registry.
Marcelo: Use the same
language as what is used by Measurement Method. DonÕt try to use some generic
language across all Measurement Methods.
Tim Carey: Have to be
concerned with role and location point. Also have to take into account where
ŅIÓ am and where the other points of the measurement method instantiation are.
Al: ippm-lmap reference path
provides reference points. ThatÕs in information model.
Tim: Yes, but itÕs needed
info when building rules.
Al: But thatÕs not a comment
to the registry.
Tim: Agreed. ItÕs a comment
to the framework. Must convey points.
Dan: Do we understand edits?
Phil: Yes.
Phil: <continues on slide
3> <discusses time topic on slide>
Dan: <at mic, not as
chair>: Comment was mine [regarding time]. ItÕs an issue about how you
interpret results and is important we are working with distributed peers. But
what you are saying is how MAs will know or estimate time. May raise issues of
scalability and other issues. But I donÕt have a solution at this time.
Phil: It depends maybe on
what sort of measurement youÕre doing? If doing a measurement randomly once a
day it may not be important. Others it may be important.
Dan: (at mic> So maybe
something that says there is no requirement on any minimum granularity of
syncronization on MAs and help people understand what kind of Methods they can
rely on.
Phil: OK. Will look at that.
<continues to section 6 text placement topic>
Al: Text needs to not just
be informational. We offered to split framework a year ago but people didnÕt
want to. We need to finish what we agreed to do and put all topics in one doc
but try to keep it short.
Dan: <as chair> Yes,
we will continue as previously agreed. We will not move section 6 to an
appendix.
Phil: So we will leave
section 6 where it is and not move it to an appendix. <no disagreement> <completes
slide 3>
Dan: <as chair> Do we
need another WG last call?
Mike: What about
environmentals? Do they belong here? How do you describe repeatable script of
tasks.
Phil: Pretty sure that framework
mentions this. And information model handles it.
Mike: But details are needed
to make it repeatable.
Marcelo: It does not belong
in registry.
Al: Maybe we need one
sentence to talk about cross traffic in framework. <Al/Dan discussion
resulting in Al agreeing to draft a sentence.>
Matt Mathis: Something jumps
out at me. There is language that says if a test is errored, you donÕt send a
report. This introduces sampling error. I want to know if it fails and why. If
suppression is in force, I want to know all measurements that were suppressed.
What if unsuppress unsuppresses just a part?
Phil: Unsuppress
unsuppresses all. You canÕt just unsuppress a part. OK. Anyway, we need more
info reported.
Dan/Phil/Matt: <agreed
that there needs to be clarification to address MattÕs comments.>
Barbara: There is a
difference between report of measurement data and report of log (what did and
did not happen in MA, which tasks were and were not run).
Dan: We will see what
changes result from AlÕs proposed text and MattÕs comment and do another WGLC
last call. Then decide if we need to run another WGLC.
Mike: It sounded like we
agreed that environmental stuff would occur in this doc. So IÕll try to
participate with Al to make sure itÕs ok.
Barbara: We will just be
saying that environmental conditions can have an impact. We will not be
providing details of when and where environmental conditions can or should be
collected or have an impact. We cannot make general statements regarding cross
traffic or simultaneous tasks that are applicable to all tasks.
Al: The statement will be
short.
Mike: When doing throughput
tests, we need to stop throughput test when there is cross-traffic because that
is a regulatory requirement when doing the throughput test for the regulatory
use case. How do you describe handling of simultaneous tasks. That needs to be
made clear.
Barbara: We must not make
statements suggesting we know what impacts environmentals have on results.
Mike: So you want to make
assumptions?
Barbara: I want to make no
assumptions.
Barbara/Mike/Others:
<trying to get clarity but not really going in a direction of resolution>
Matt: There needs to be a
wrapper or shim that wraps the tool and enforces different conditions.
Dan: We wonÕt resolve here.
I encourage you to try to resolve before leaving Toronto. We will see what is
in the next draft. If there are any remaining comments, please send them in the
coming week (by July 31). Thank you.
----
Brief discussion on Use Case
draft comments (at 52:50 in mp3 recording)
Benoit Claise: In use cases
I saw impact. I reviewed use case doc and wasnÕt sure whether user initiated use
case was in or not. I found in framework (section 5.6 – items beyond
scope of initial work). It would be nice if that language that itÕs not in
scope were in use case doc. Need to reference In use case RFC on monitoring.
Dan: You will send comments
to list? Thank you.
----
Information Model (at 55:25
in mp3 recording) (http://tools.ietf.org/agenda/90/slides/slides-90-lmap-2.pdf)
Everything is a task – this makes things simple
Jrgen Schnwlder: Who has
looked at -01? <some hands raised>. Trevor us not here. He has done most
of the work and I am representing it. <slide 2> Everything is considered
to be a task. <slide 3> More about tasks. <slide 4> Configuration
<slide 5> Channels <slide 6> Suppression <slide 7> Reports
<slide 8> Miscellaneous <slide 9> path forward?
Dan: <chair> What do
you suggest?
Jrgen: Suggest creating a
rev that addresses clarifications and minor edits and proceeding to WGLC.
Al: Is there more to do
regarding roles and measurement points?
Tim: I havenÕt had a chance
to model -01 in BBF, so canÕt say yet. Question about chaining.
(Tim has aligned BBF doc with -00 version. He will look at -01 to see if
Roles idea.)
Jrgen: In scheduling there
is concept of output of one task is input to another.
Concept of
chaining tasks - please elaborate
Concept of roles needs to be added
Tim: OK. I will be creating
new BBF model by September.
Phil: Note roles are needed.
Dan: <chair> Maybe we
should go ahead and spin off a version now and start a last call in August on
the -02. Have a 1 month WGLC. OK? <no objections>
----
BenoitÕs
feedback (as AD on use cases
S5.6 of
framework – copy across – to clarify end use case is out of scope
Add a mention of the pervasive monitoring rfc
----
Large MeAsurement Platform Protocol
based on HTTP (at 01:12:30 in mp3 recording) (http://tools.ietf.org/agenda/90/slides/slides-90-lmap-1.pdf)
Will there be a separate data model
draft? No.
Now everything is a task, management things (send report, ..) need to be identified and included in schedule – do through registry
Marcelo: <slide 2> <slide 3> <slide 4> <slide 5> <slide 6> <slide 7> Data Model
Dan: Encourage you to think about structure of data model and make it as modular as possible.
Benoit: Wonder if solution isnÕt YANG. In Sunday tutorial we learned how to get JSON out of YANG.
Marcelo: OK <slide 8> <slide 9>
Dan: Is it open-source? Can it be referenced and looked at?
Jrgen: No. Not released, but you can ask us
Dan: OK. So for time-being it is closed.
Marcelo: Still work to be done. Closely aligned with what Samknows platform does
Tim: Going back to Report protocol <slide 6>. Note you are using post. Was there consideration of using a uniform contract? Just trying to understand if you considered it and decided not to do it.
Dan: <at mic, not chair> There is more than one proposal. This is a lightweight proposal using a minimum footprint. If there is something in framework that is not correct, then that needs to be addressed.
Tim: My concern is that now I make assumption on lightweight element, I feel uncomfortable that I donÕt know what it looks like in a more constrained environment.
Dan: So it may not be constrained enough. OK. Note we may end up with more than one protocol. So different environment such as more constrained may need different protocol. charter doesnÕt limit us to one protocol
Mike: You shouldnÕt forbid expansion of this and other implementations should be allowed to expand on it.
J. Ott: I wouldnÕt worry so much about implementation complexity.
Jrgen: There are no home routers that cannot do http. For other proposed MAs too. , lwig constrained devices are not MAs
Tim: Yes we have definitions of constrained devices. I just wanted to make sure that we were precluding certain types of constrained devices. I have no problem with our precluding certain types of devices. I just want to make sure we know weÕre doing it. I do want to constrain the number of protocols that come out of this.
-----
REST Style Large Measurement Platform Protocol (at 01:34:00 in mp3 recording) (http://tools.ietf.org/agenda/90/slides/slides-90-lmap-5.pdf)
Dapeng Liu: <slide 2> <slide 3> <slide 4> <slide 5> <slide 6> <slide 7> <slide 8 - 14> <slide 15> Running code and deployment.
Dan: <at mic> These are not components of the LMAP architecture.
Dapeng: Correct. <slide 16 - 19> <slide 20>
Al: We had heard measurement traffic was illegal in China.
Dapeng: Not aware of this.
Al: I encourage you to investigate.
Lingli Deng: This [measuring] has been going on for several years and I donÕt know where your information comes from.
Jrgen: It seems in your implementation the Controller connection has to get through the NAT. MA is http server – so how talk to MAs behind firewalls? How do you establish tcp connection inbnd
Dapeng: Good point. We can get the mapping between public and private IP address. And there are other ways like Dynamic DNS.
Jrgen: You should document how you work through the NAT. Another question – can you describe how you do partial updates?
Dapeng: The URI design could be flexible. Right now there is a single command that is used to update the device. No partial updates.
Tim: I think you are missing ability to do partial updates. I think it is needed. Especially if you only want to send Suppression and nothing else. missing ability to do partial updates, eg suppression. So should use post rather than put. This is mostly about configuration. Reporting is ok.
Dan: What are your plans for further update?
Dapeng: A few weeks after info model is up we will make an update.
Phil: Do you have plans to release the code?
Dapeng: CanÕt answer that now. Co-authors are from other companies.
---
<Moving on to broader discussion on protocols> (at 01:50:22 in mp3 recording)
Dan: There is one more contribution announced on an email. Arne?
Arne Oslebo: We are planning on submitting a different protocol proposal. Yang is a good option, weÕre are writing a draft. Rest is good, so will be sconf not netconf
Dan: So estimate is that in a couple of months we will have 3 protocol proposals and maybe more. We will then start work on protocol selection. Note that charter does not require us to restrict to one protocol.
Al: I think charter says ŅTheÓ protocol. I thought we should allow multiple but was shouted down by those who wanted just one.
----
Liaison from BBF and proposal for interim meeting in Dublin (at 01:54:00 in mp3 recording)
Dan: There is BBF meeting in Dublin the week of Sept. 15 and there are proposed joint activities. The meeting would need to be open and allow for remote participation. We are in discussions. Before we see who could make it we need to see if people think there is benefit.
Dave Sinicrope: There will definitely be an invitation extended from BBF for the week in Dublin. We need to work out logistics on remote participation. Once we get details worked out, we will send invitations. On another point there was a liaison from BBF with WT-304.
Dan: What will be process for comments?
David: liaison points to WT-304 text, asking for review – send comments on lmap list BBF participants will monitor lmap list for comment discussions.
Dan: This is not related to WG charter and list is only supposed to be used for chartered items.
Dan –
AD advice on whether we can use lmap list for this
purpose?
Benoit – better to keep simple than go through formal liaison
David and Barbara: Working with other orgs is specifically in charter.
Dave
– still finalising but will extend invitation
to non-memebers to Sept meeting
Benoit: Correct. This is an ok use of lmap. The proposed process sounds simple so I am in favor.
Dan: OK.
David: But wait till next liaison. There will be another liaison that has the official straw ballot version that will be posted in a few days.
Dan: What about telling BBF about WGLCs?
David: It would be ok to send liaison to BBF letting them know of WGLC status and telling them to watch lmap list to know about last calls. This is just to hit people who are not currently on the lmap list. Regarding guests at BBF: As we know BBF is a membership organization. For the Dublin meeting, BBF will be extending an invitation to non-BBF members who participate in lmap, and will likely waive the registration fee. There are several sessions they will be invited to attend.
Dan: Maybe you can tell us more about the workshop?
David: There are thoughts that one of the session would be educational about YANG and modeling. - maybe Juergen could do
Juergen – what will we do in interim? Dan – info model last call comments resolution. Review protocols. Protocol selection – how to do.
Dan: LMAP specific part would be focused on chartered work?
David: Yes.
Phil: There will be formal LMAP interim to work on info model but also to look at next steps and what work might be done elsewhere.
David: BBF leadership will work with LMAP chairs.
Phil: So LMAP interim on Monday afternoon and WT-304 on Tuesday?
David: I can communicate that back.
Jrgen: I have concerns regarding my ability to attend. I was also hoping we could finish the info model before then. I was wanting to work on protocols at an interim.
Dan: My thinking is that we need to make progress on the protocols. We could start discussing criteria for protocol selection. We could also discuss how protocols will be maintained..
Jrgen: And a reason I really want to know the agenda is because there may be a different interim scheduled for me that same week.
Benoit: There needs to be at least one chair. Need to know who would attend. How many BBF guests might there be? <3 hands raised> This is not excessive. Who thinks having interim is bad idea? <no hands; but depending on agenda>
Dan – straw poll – maybe 10 hands
Marcelo: It definitely makes no sense to go without agenda.
Jrgen: For me info model is mostly done and I donÕt think there is much work left.
Dan: I am more cautious. Vacation time is coming and there may be enough comments that there may need to be another WGLC.
Dan: I think there can be good discussion on criteria for protocol selection. Anything else on interim?
Barbara: Just wanted to mention that liaison from BBF also included the TR-069 data model created from the -00 info model draft.
Benoit: Interim needs to be announced 30 days in advance. So we need to finalize it soon.
Dan: Any other questions?
Brian – ippm registry milestone – do lmap milstones impact ippm? Think weÕre ok, data model milestone is the thing that might be impacted
Jrgen: Info model may be done prior to December.
Dan: No rule against finishing early.
Brian Trammel: No relationship between lmap and ippm milestones. Is there any registry dependency?
Dan: Data model. Now is open microphone time.
Lingli: WG may also be interested in looking at additional use case that was submitted.
Dan: The draft refers to items out of scope of initial chartered work but may be considered for next phase of lmap. Considering that in the meantime is fine.
Lingli: We can do this the hard way or the easy way. Coordination can be discussed in abstract or specific.
Dan: out of scope of first phase of lmap, but Can be considered for next phase.