[nm-wg] summary of last call, the next call, the next f2f meeting

[Leese, MJ (Mark)]

Hey folks,


The next call
-------------

Is this Tuesday (the 21st) at the usual time (8am PST, 10am US CST, 11am US EST, 4pm GMT, 5pm CET etc etc). Dial-in info:

	(Inside the U.K.)  0871 271 2900
	(Outside the U.K.) +44 (0)870 050 6800

	Enter the meeting code 255404 when asked.

	U.K. users are charged at 8.51p per minute (plus VAT) per minute.
	Non-U.K. users will be charged according to their local telecom providers.

Agenda:

1. Continue discussing the draft developers guide (http://www.cis.udel.edu/~swany/nmwg/devguide-05-dec-04.pdf) for the new/"normalised" schemas from Dan, Jason and Martin.
2. Review any progress Mark and Paul have been able to make stabilising the current schemas.
3. Finalise date for next face-to-face meeting, so people can book flights.


The next face-to-face meeting
-----------------------------

This is going to be a half day around the next Joint Techs meeting, in Salt Lake City, Sunday 13th to Thursday 17th February.

After the last call a meeting on the morning of Thursday 17th looked likely. This is now doubtful, as a lack of flights Thursday afternoon would leave people stuck in town until Friday morning. The two alternatives suggested are:

* Wednesday evening - ready for people to fly out Thursday morning.
* Sunday morning - due to the available flights, most people will fly into town on Saturday ready for 1pm meetings on Sunday. So they'll be in town, and free, Sunday morning.

If you can't make tomorrow's call, and you'd like to attend the meeting, please let me know your preference asap. A final decision will be made during the call, so that people can book their flights.


The last call
-------------

Notes from the last conference call are below. Please let me know if Susan and I have made any mistakes. Thanks to Susan for also taking notes.

Five point summary:

* We like the new schema work a lot, and it's something we want to pursue
* Martin and Jason are working on a simple, sample Perl implementation to look at. Changes to the document will proceed in parallel.
* We will look at producing sample implementations in several languages to make it easier to adopt and therefore promote uptake. We'll start this before the proposed face-to-face meeting in February, so we can discuss problems encountered/lessons learned at that meeting.
* We want some kind of representation at the next GGF (Seoul in March) to promote our work in Asia-Pac

Actions from the meeting (for Dan, Mark, Martin and Richard) are at the very end of this email.


Cheers,

Mark.




NMWG Conference Call: Tuesday 7th December 2004, 16:00-17:15 GMT
-----------------------------------------------------------------
Present: Mark Leese (Daresbury Laboratory); Jeff Boote, Susan Evett and Matt Zekauskas (Internet2); Dan Gunter and Brian Tierney (LBL); Paul Mealor (UCL); Martin Swany (Uni of Delaware); Richard Hughes-Jones (Uni of Manchester)


Agenda:

1. Discuss the draft developers guide (http://www.cis.udel.edu/~swany/nmwg/devguide-05-dec-04.pdf) for the new/"normalised" schemas from Dan, Jason and Martin.

2. Discuss when to hold the next face-to-face meeting. The three current suggestions are a half-day fitted in with the next Joint Techs meeting, Salt Lake City (http://jointtechs.ornl.gov/SLC2005):
Sunday February 13th, PM
Wednesday February 16th, PM
Thursday February 17th, AM

Also discussed:
3. the next GGF.


Contents:

1. Draft Developer's Guide:

Mark asked Dan and Martin to summarise give an overview of the document. This resulted in a section by section discussion.


Section 2: Base schema

Dan: The main point here is that we're separating out data and meta data. a request would contain meta data only, while a response would contain meta data, real data, and some identifiers to associate the two with each other.
Mark: Can we put what you've just said, this general description, in this or another NM-WG document? This would be for the benefit of people who aren't as familiar with the work as we are:
ACTION: Dan/Martin to put general description of purpose of these schemas in the document (e.g. separating data from meta data) for the benefit of people who aren't that familiar with the work.

Martin: It should be noted that extensions aren't done in a typical OO way. It's the base stuff OR the extended stuff.

Rich: Do the URLs in the namespaces resolve to anything. Is there anything at the end of these URLs?
Dan: You don't have to have something at the end. Some people make them do that, but it's not essential. The namespace just has to be unique, which is why people often use a URL.


Section 3: Examples

Martin: The examples are designed to demonstrate the differences between data and meta data. We also have examples for ping and iperf they just need updating to the latest format.

Jeff: The response in the traceroute example doesn't show the requesting program being given an idea of how many results to expect. Is it expected to read until the response stops/runs out or will it receive a notification of how much to expect? This was an issue we addressed (albeit in a different way) in the current schemas.
Martin: This is possible but it's just now shown. I'll include an example that does.
ACTION: Martin to include an example in Section 3 that shows the requesting program being told many results to expect.

Martin: We've also been thinking of having the main examples, and then picking one of them to provide extended examples of, that show various permutations.

It was agreed that we'd concentrate on the main examples first, but Martin would insert placeholders into the document about permutations. This will ensure we don't forget, and will let people know they're on the way.
ACTION: Martin to add placeholders into the examples section to indicate that we will take one main example, and provide various extended examples showing possible permutations.


Section 4: Extensibility

Dan: The extensibility section aims to show how the schemas could be extended to meet specific testing needs. However, it's a very simple example, and it may not be entirely useful.
Mark: It's good to have a simple example, to get the fundamental points understood before doing more complicated things. A more complete, complex example can always be added later.
Paul: Yes, this is somewhat less "practical" than the examples in previous sections, but it gives users an idea of how to get started. People could just use the finished examples, as they cover the most common tests (ping, iperf etc.) but this shows them how to create their own.


Section 5: Use with Web Services

Dan: Was this useful to people? I haven't had chance to test it yet, but it's close.
Rich: I couldn't reproduce it now, but I understood most of it as I was reading.
Mark: It's very important. This is a developer's document after all.

Dan: I'm thinking of taking it further, e.g. "We're using this tool. We do this to generate these stubs. We then do this..."
Jeff: Yes, a complete example would be really useful. Loads of us are new to all this (Web Services) work.
Rich: I second that.

Martin: The document talks of generating code from WSDL. I think we also need to represent the code first approach. Not just for completeness, but because there's a lot of code already out there, and because its particularly applicable to SOAP::Lite implementations.
Mark: I'd go further than this. I think we need some sample implementations, maybe Perl, Python and Java. We need to present the work with example implementations, so that people really have no excuse for not adopting our work.
Dan: Well I think I could provide some Python, but on the "code first" issue people should know that they still need to understand how the schema works a bit before they begin if they are going to interoperate.
Jeff: Internet2 could do a Perl example.
Mark: Okay, I'd hope that Europe could provide the Java.
Dan: Microsoft is the 800lb gorilla here, so we may need .Net as well.
Rich: This is an area where we could get Geoffrey Fox's group involved if we feel we need some technical help, e.g. no NM-WGers are familiar with .Net. We should provide help or examples and let Geoffrey's group either hold a workshop or bring NM-WG people on up to speed or provide examples themselves.
ACTION: Richard to follow up the issue of assistance with Geoffrey Fox (especially .Net).
Martin: GSoap (stylised .h files) could be added to the list if we wanted a C example.


The call was then opened up for general question on the developer's doc.

Richard: How does the minutiae of the work we've already done, e.g. hours of discussion over what time formats to use, fit into this?
Martin: By providing base or extension types that are valid/compatible with what we've already decided in terms of the old schemas, the hierarchy document etc.

Mark: Are we still planning on NM-WG being gatekeepers for groups wishing to define extensions to the schemas?
Martin: Yes. If you came up with something new, you'd put it in an organisation specific namespace and mail it to NM-WG. NM-WG will eventually vett it, presumably approve it, and put it in the NM-WG namespace. While that happens you can use it, but with your own namespace. This would allow people to get work out quickly, whilst guarding the integrity of the NM-WG namespace.
Dan: We'd have to be responsive on this, so we don't get different groups using/suggesting similar things in their own namespaces. The flip side is that we shouldn't let in things we're not happy with, just because we're under time pressure.
Martin: Yes. We also need to be careful that we don't get two different ways of expressing the same thing. We need one canonical way of doing each particular request/response.

Jeff: Are we looking at eventually producing one big schema from all the small ones, or keeping them separate.
Martin: The latter.
Mark: Tied into this is my Devil's Advocate question. What would you say if I was nasty tool developer ;-) and said "Oh, I don't like this new work. I don't want to deal with lots of different schemas. It's better when there's just one schema that does everything, even if it's not as powerful".
Martin: If you wanted to, you could put all the components together into a single one. I also think you could perform stylesheet translations between the old and new work. 
Dan: People should really want this. In general, they want to focus on what they measure. The new work will give them the opportunity to build schemas that will fit them like a glove. No one should care about the whole thing. It should be easier to pick and choose subsets. If versions change, you don't have to change everything, and it's easier to stop people requesting things that don't make sense.
Brian: Presumably, if everything is in components, you can have code associated with those components (decomposable and exchangable). So you could jump to a library of code if you were using a particular schema component, and not have to worry about generating anything from WSDL?


Next steps...

Mark: So how long before before we're happy to start an example implementation? How easily can a system be built to implement this? Is it something we can do as a subgroup, or do Dan, Martin and Jason have to work on it locked away in a room?
Martin: I'm already working on a perl implementation with Jason. It'll get sent out as soon as it's half baked. Before Christmas we hope.
Dan: This can be fed into the document..."If you want a Perl implementation, then do this..."

Rich: So if things go to plan, we could present some experiences at the proposed face-to-face meeting.
Jeff: I don't know how fast my stuff could be produced. It may be close, but is unlikely to be done that fast.
Mark: We'll have a better idea in January, but it definitely sounds like we're moving on this.
Rich: There needn't be code by the meeting for there to be things to discuss. If people have made progress we'd want to know about the problems found on the way, and the lessons learned.




2. The next face-to-face meeting:

The suggested dates were:
	Sunday February 13th, PM
	Wednesday February 16th, PM
	Thursday February 17th, AM

People very likely to be able to attend are: Dan, Eric, Jeff, Richard and Matt Z. Brian is pretty sure he can attend. Martin would like to, but would prefer Sunday to avoid having to miss classes he's teaching. Paul will have left UCL by this stage. Mark would prefer to join this by phone, and use the airfare for GGF in Korea the following month.

In terms of the day, it was slightly dificult to decide, mainly because the programme isn't advertised until 3-4 weeks before.

Wednesday was ruled out because people think there will be other meetings happening that they would perhaps like to attend. In particular Richard mentioned an email from Heather Boyles advertising a "Passive Monitoring" talk from Joerg from NLANR, and an "Intruder Workshop" from Gregg at Indiana, for Wednesday afternoon.

Sunday morning was possible, but not the suggested Sunday afternoon, because people thought there would be tutorials to attend. Thursday morning was the most attractive in terms of being able to attend other meetings. It's towards the end of the meeting, and people feel it's a time when something new won't be added to the programme.

ACTION: Mark will check with Eric to see if we now have enough consensus.

NOTE: that the situation has changed since this phone call. Read the top of the email for the latest info.




3. The next GGF:

Richard asked if any decisions had been made about the next GGF (13), which is in Seoul, March 14-17 2005.

Mark has mentioned this to Eric. He thinks it's important that NM-WG has some representation there. The group is very US and Euro centric, and he thinks we need to make some effort to get Asia-Pacific visibility, buy-in and active involvement. The consensus of regulars so far is that Seoul is too far/too expensive to have a full NM-WG meeting. One suggestion is that 1-3 of the "core group" go to Seoul to demo some schema implementations and present the new schema work. If we have a good turn out then excellent. If not, then we have at least made the effort!

Mark is currently okay to go. Richard has funding. Brian is possible. Dan thinks he's unlikely to attend. Jeff will discuss this with Eric to see if someone from Internet2 can attend. Martin would like to attend, but will have to look into it.

This can be discussed further in the New Year. 




Actions:
--------

* Dan/Martin to put general description of purpose of these schemas in the document (e.g. separating data from meta data) for the benefit of people who aren't that familiar with the work.
* Martin to include an example in Section 3 (Examples) that shows the requesting program being told many results to expect.
* Martin to also add placeholders into the Examples section to indicate that we will take one main example, and provide various extended examples showing possible permutations.
* Richard to follow up the issue of assistance with Geoffrey Fox (especially .Net).
* Mark will check with Eric if we now have enough consensus over when to hold the next face-to-face meeting.