[nm-wg] Notes from last NMWG call

[Eric Boyd <eboyd@internet2.edu>]

Attendees
Eric Boyd, Tanya Brethour, George Brett, Jeff Boote, Dan Gunter, Loukik 
Kudarimoti, Mark Leese, Paul Mealor

Agenda walking through request schema from last meeting

Items from last call:
ACTION ITEM: Eric to make request today via the GGF website for a 1-day 
slot for the GGF NM WG in Belgium, noting on request that it should not 
overlap with the NMA meeting.
ACTION ITEM: Mark & Dan will assign to someone (or work on themselves) 
creating presentations on the R/R schemas.
ACTION ITEM: Eric to ask Iosef if he can attend GGF meeting (Yang cannot 
attend).
ACTION ITEM: Yang to request that either Iosef attend GGF meeting or that 
one of the two of them create a presentation Eric can give on the results 
of the Request schema fetching tests for MonaLisa.
ACTION ITEM: Dan to send out notes on what he'd like to talk about in Belgium.
ACTION ITEM: Dan & Martin will have a more concrete example of the schema 
to share in Belgium.
[] (note: this is corrected below): Get rid of the sentence indicating that 
they can be used in conjunction - command line takes preference.
ACTION ITEM: Mark to rewrite including: a) Item 2 is legal, b) Item 3A is 
legal, c) no mixing of 2 and 3A, and d) respond "I don't support 3A". Add a 
note that we recognize that not everything can be covered under this but 
too bad and give number 8 as an example of something that could only be 
done w/Item 2.
ACTION ITEM: Mark to implement changes and to work all the discussions into 
a new version of the requirements. Mark to get new version out before next 
call on 10 August 2004.
========
Eric - go thru action items and then dive into request schema.
ACTION ITEM: Eric to make request today via the GGF website for a 1-day 
slot for the GGF NM WG in Belgium, noting on request that it should not 
overlap with the NMA meeting.
Have requested, as soon as know what people's travel is.
8 hours one day or two 4 hour slots.
Iosef will be attending, interested in seeing if he can ger ReqSch between 
MONaLisa adn Pipes
Tanya -- no progress made yet, will know more next week.
ACTION ITEM: Mark & Dan will assign to someone (or work on themselves) 
creating presentations on the R/R schemas.
Still uncertain?
Mark -- can say something about request schema
Dan -- can say something about response schema
Eric - give talk on presentation talked about last time.
Mark - trying to get Martin to do it
ACTION ITEM: Eric to ask Iosef if he can attend GGF meeting (Yang cannot 
attend).
Iosef can attend
ACTION ITEM: Yang to request that either Iosef attend GGF meeting or that 
one of the two of them create a presentation Eric can give on the results 
of the Request schema fetching tests for MonaLisa.
ACTION ITEM: Dan to send out notes on what he'd like to talk about in Belgium.
Dan sent those out.
ACTION ITEM: Dan & Martin will have a more concrete example of the schema 
to share in Belgium.
AI (note: this is corrected below): Get rid of the sentence indicating that 
they can be used in conjunction - command line takes preference.
ACTION ITEM: Mark to rewrite including: a) Item 2 is legal, b) Item 3A is 
legal, c) no mixing of 2 and 3A, and d) respond "I don't support 3A". Add a 
note that we recognize that not everything can be covered under this but 
too bad and give number 8 as an example of something that could only be 
done w/Item 2.
Mark just sent them out to the list.
ACTION ITEM: Mark to implement changes and to work all the discussions into 
a new version of the requirements. Mark to get new version out before next 
call on 10 August 2004.

Eric -- Folks check email and see doc Mark just sent out.
Mark -- Not much use going thru all the points again. Scroll to bottom to 
see outstanding issues.

Outstanding Issues
Outstanding Issues Some points from the "Outstanding Issues" section were 
also discussed earlier in the day, during Paul Mealor's "walkthrough of 
request schema" session.
1. Schemas for interim communication (that between initial request and 
final result message): It was recognised that WSRF may provide solutions 
for much of the envisaged interim communication (e.g. requestor and 
monitoring system negotiating over the running of on-demand tests). 
However, it was also noted that WSRF is not likely to be available as a 
polished product for some time, and in the interim some "roll your own" 
solutions may be required to progress, even though they could be temporary 
(throw away) measures. So, interim communication is still an outstanding 
issue, but the point should be updated to show that we have considered
Mark -- Hope WSRF will answer a lot of the issues here. Roll you own 
solutions.
Dan - group at LBL just released a WSRF. Moving quicker than expected.
Erc - is this point different than what's expected as response to request?
Mark - sort of. Predictable get back last resource. Doesn't 1114
60 sec Iperf - response - no you can't have 60 but 10
Haven't considered between
eric - rather than answer question.. we are or are not doing WSRF -- just 
see if figure out the state machine, the programmic responses possible. To 
get implementation guide written, see where at re: WSRF
Mark - seems fine
Jeff -- is WSRF language for describing limits? I want make request and see 
what limits are. If not best - Yes, No, Not Now. "Not Now" even difficult
Tanya - seem that endless parameters could be changed for each tool This 
could be ver complicated.
Mark -- idea of negoatiation stuff will be complicated. Some people want 
discovery mechanism. "I'm person X, what can I do now?" Do some kind of 
discovery..
Jeff -- agree with capability discovery should be outside of response and 
request schema
Mark -- agree discovery capability should be outside. Send me request, I'll 
be happy to respond when able to - in 10 min time or such. Not sure what 
instrument communication will be.
Jeff -- have thought about it a lot. Trying to figure out how to do more. 
Added "not now, I'm busy" message. Logic in our stuff with event horizons 
-- have happen between start and end time if not then a busy signal. If 
ouside that event time then try again later. Be real careful with this.
Mark -- please send thoughts out to the email list/
Jeff -- sure [] Jeff send thoughts on #1 to the list.
?? -- interim communications - looking -- one way delay from these routers. 
All these measurement boxes. Don't know which to choose, can you tell me 
which to use? A possible example. Currently, make it ad hoc. User want to 
be in control. Does this make sense?
Dan - makes sense, but going to be hard to do.
?? -- yes and no not very effective. Has to be more than just yes and no.
Dan -- this needs to be discussed more in detail
Mark -- see what comes out in email between now and Brussels.
[] talk about this in Brussels... add this as agenda item for Face to face.
Dan?? -- big issue relates to default values that are filled in.

ITEM 2. New requirement: Yes, the request schema should support the ability 
to specify the return language and transport mechanism for responses to 
requests. This would allow alternatives to the NM-WG publication schema to 
be used.
Mark -- make request schema lots more useful. give last 5 values rtt. Paul 
knows -- don't think we can do this with current schema.
Eric -- to do this move in direction Dan and Martin talked about last week.
Dan -- grouping schema might do it
Eric -- pretty rradical overhaul. Should stabilize on what have now and 
consider this for version 2
Dan - ok -- some feature creep.
Eric -- some point capture document for GGF and then move on
Dan - long process for even informational document in pipeline. Took DAMED 
doc two year.

ITEM 3. Sharing of common components between request and publication 
schemas: We'd like to do this, so no change.
mark -- sharing components between request and response schema. Didn't get 
too far with it.
Eric - aim for by Brussels. This version is pretty much fixed. If want to 
make progress over last meeting. Make first stab at breaking out components 
and present at Brussels.
Dan - create side by side comparison in spreadsheet. Flatten out to see 
clearly, quicker
Eric - who?
[] Paul and Mark could work this up with help from publication schema 
person. Dan said he could help too.
[] Dan first cut at spreadsheet setting up layout. Mark & Paul second cut. 
anyone else can help. Dan will send to the list.
[] Eric -- aim for by next meeting Dan touch once and then Paul & Mark will 
touch it.
Mark - next call wont be there.
Eric -- move out the next call?? 3 yes'es. Dan, Paul, Loukik,
Would one more week help? Mark - yes Dan, paul, Loukik yes
[] set next call three weeks from now.
[] Eric will send out reminder.

ITEM 4. Should the schema support one or more requests at a time?: It was 
decided that the schema should support multiple requests in the same 
message, so this point can be removed from the requirements.

ITEM 5. Allowing wildcards for characteristic names e.g. "path.delay.*": 
Still to be discussed.
mark - possibility to specify wildcards. path* show one way and path. Where 
folks can use wildcards and to what level they can use them too.
eric- counter the * becomes a query, a characteristic query. If give ??* -- 
give a path request. Implicit conclusions
Mak - path.delay* - if system had data for one way and round trip you'd get 
both
?? - wouldn't that make it more complicated?
Mark -- yes - but what we need to decide, what's lesser of the evils?
Jeff -- if asking for more that one thing at once. Makes sense with 
historical data, but not when running multiple tests all at same time. 
Makes things complicated... Difference between request to do test versus 
request for historical data. Gets more complicate when asking to run tests.
Tanya -- how complicate?
Jeff--- question of which test are being requested. All tests, the entire 
list. If non destructive that's ok, but some test interfer with each other.
Tanya - shouldn't it be back end responsibility?
Jeff - which test they want? How determine which one they really want / need?
Dan - have some way to query what characteristics can be requesteed. Wild 
card becomes two queries.
Jeff - first a capability request and then the tests
Dan - make a request, give me your characteristic names
Tanya - should have that.
Eric -- is there consensus?
Jeff - fine with what Dan suggested more than letting system decide.
Eric - puzzled - if make request... give me possiblities, use wildcard as 
part of that.
Jeff - just give me all possible characteristics
Eric - no wild card
Jeff - right.. user doesn't have to know.
Tanya - can user request set of metrics or one by one?
Jeff -- depends on how presented to requester. If done one at a time
Eric - answer you question?
Mark - no. Are we allowing wild cards or not
Dan - no wild cards, different sort of request
Eric - outside schema? put on the to do list?
At present wild cards are not allowed, but may be used later in capability 
discovery (1142)
Dan - think you'll nee capability discovery immediately, but will be neede 
later at the leat.
Eric question....
Dan -- almost same - give me all the characteristics.
Eric - separate flag for characteristic requesting?
Mark - implicit built into the repsonder... there's a star, know what I can 
measure or deliver.
Eric - just a small change
[] Mark put in and explain what it means. Remember when designing how to 
implement document.
Dan - repond with characteristics and no data
[] Look at what will have to change to make proposal to that.
[] Mark add some text
Jeff - when do capability discovery have to have more in line with DNS 
model or JINI Services model when looking for certain capabilities.

Tanya -- only one request at a time?
mark - no multiple requests in one message now
[] Eric will work on #5 before Brussels mtg.

#6
Mark -- predefined at source or destination
[] add that in ???
Dan - confused - new option looking at?
mark - don't specify source or destination, you follow both ends.
Jeff- speficy port both times
Mark - just use port once without attributes
jeff -- up a level -- is there some guideline to make a new tag or existing 
tag? Is there good line to make these decisions in general?
Mark - guess so
Jeff - what's the advantage?
Mark - advantage is - if different from source and destination you have 
different source and destination tags. Whereas one port tag with one 
attribute you'd only need one set of tags.
Jeff -- general to simplify or not
Dan -- attribute is more flexible. Port=foo and that works.
Eric -- Wrap up now -- 3 weeks go thru 7 8 9 and review first cut of common 
components and response schema
[] next call review 7,8, 9
mark -- next call have quick 5 min chat about use cases
[] use cases about request and publication schemas used - not formal
Eric -- Mark send out request and clarification to list
[] Mark will send out request for use cases to mail list.


Eric L. Boyd
Performance Architecture & Technologies
eboyd@internet2.edu
734-352-7032 (work)
734-834-6708 (cell)

Office 230
Internet2
3025 Boardwalk
Ann Arbor, MI 48108