20090417 arb OOC Minutes

From HL7Wiki
Jump to navigation Jump to search

Architecture Board

April 17, 2009


Call to Order

The meeting was called to order at 9:00am with John Koish as chair and Tony Julian as Scribe.

Attendees

Name PresentWith AffiliationE-main address
Curry, Jane Yes ArB Health Information Strategiesjanecurry@healthinfostrategies.com
Davis, Brian Yes Guest3rd Millenniumbdavis@3rdmill.com
Duteau, Jean-henri Yes MnM GPIjean.duteau@gpinformatics.com
Jorgenson, Don Yes ? Inprivadjorgenson@inpriva.com
Julian, Tony Yes ArB Mayo Clinicajulian@mayo.edu
Koisch, John Yes ArB 3rd Millenniumkoisch_john@bah.com
Mead, Charlie Yes ArB Booz Allen Hamiltoncharlie.mead@booz.com
Mulroney, Galen Yes ? VHAgalen.mulroney@vha.gov
Nelson, Dale Yes ArB II4SMdale@zed-logic.com
Ocasio, Wendell Yes ArB Agilex Technologieswendell.ocasio@agilex.com
Parker, Craig Yes MnM ASUCraigparkermd@gmail.com
Quinn, John Yes ArB Health Level Seven, Inc.jquinn@HL7.org
Shakir, Abdul-Malik Yes ArB Shakir ConsultingShakirConsulting@cs.com

Don Jorgenson: Alpha project: Would like to see alignment with HITSP over the nexe 90 days. They have security/privacy, but will have to rework. To capture a specification down the stack, at what level does it become technology, moving out of engineering. Low technology is needed to interoperate.

Don Jorgenson: SPecs for encryption (TLS) , message level encryption: diving to the bottom, what is the scope, and how do you handle. Where do the specs get captured?

John Koisch: service, collaberation

Don Jorgenson: documnent to repository to service. Document is parameter, has target. collaberation level resolving target address, details of the support/discovery, then we build a soap, with ws-secure, users, patient, declared role. Down to message, must bind to transport

John Koisch: There are other details. single interface - send_document. There is an implementation, verifying data as necessary, doing the business level stuff. Logically there is another guy who takes care of the business stuff -separation of concerns between validation and business level. There are a couple of services. There is a need at some level to validate and specify the service, and security components. You may define integration points - topic specifications. You can build it at a high level.

Don Jorgenson: Determining what is compatable with the other ends capability. Low-level technical stuff needed to specify interoperability. Below platform-independent there is other stuff.

John Koisch: several potential services. There is a collaberation specification. At the high level, you call the encryption service.

Jane Curry: bring in engineering viewpoint.

Wendell Occasio: not in engineering?

John Koisch: Expression of quality of the interaction.

Don Jorgenson: taken as a whole it has a profile, that defines end-to-end. I am concerned about lower level stuff. How do we capture it?

Jane Curry: in IHE profiles it is platform specific.

Don Jorgenson: we specify 256 bit encryption

Wendell Occasio: that is not platform

Don Jorgenson: you can do it on a platform

John Koisch: OMG and the community think of platform as a well defined stack. We dont acknowledge that they have all the platforms right. You are defining your own platform. NCI platform specific security spec is platform defined - they define there own stack of stuff -CABIG security platform.

Don Jorgenson: we need to communicate the information.

Charlie Mead: There is a one to many mapping from the top down, that is why bottom-up is hard. You have to distinguish between platform and technology binding.

Don Jorgenson: I dont see how I can stop until I get low.

Charlie Mead: thats right. Specifications happen at the implementation level.

Don Jorgenson: below that you have technology - what is the deliniation.

Wendell Occasio: tls with 509 certificates is platform specific, but if you say you are using APACHE, that is the technology binding. Platform specific is JAVA, TLS. At the platform-independent we can discuss what we will do. Higher up we say you have to use security on the wire. RM-ODP calls it a stub. the part in the channel that provides the encryption.

John Koisch: the problem (our fault) is that CABIG things about the engineering viewpoint, and ignores the others. Charlie said "You need the engineering" - you are defining a topic specific - IHE ... platform model for security. CABIG only has the engineering - has to be redone to provide CABIG topic security model. Saying TLS and 509 is at the platform specific computational and engineering.

Wendell Occasio: you have to show how it binds e.g. web services/SOAP/https.

Don Jorgenson: you say web services which in OMG speak is a platform.

John Koisch: you platform specific is my platform independent. The trick is to get the spec right, to document why you are using x509.

Wendell Occasio: alternative - thats one way, wire encryption, different from encrypting payload and digital signature. You might want to say the security at the transport level, even if broken, you want to secure the payload.

Abdul-Malik Shakir: Don was asking for a platform specific e.g. project on a screen, implementation is Dell computer on Sanyo projector.

Don Jorgenson: technology

John Koisch: RM-ODP talks to it.

Wendell Occasio: The technology viewpoint. It is the lower level.

Don Jorgenson: this is helpful

John Koisch: Technology can bind through the abstraction layers to anything. You can bind to conceptual level. You are building something - the stairway to heaven covers that at least you are binding conceptual to technology. If you specify the libraries and specification stack - security common model security stack. We can model it.

Charlie Mead: if you have all the cells built, and have traceability built, you can only have direct binding and be done, but it is not complete.

Don Jorgenson: profiles take a thread and take it all the way down.

Jane Curry: you can have multi-layer in the same cell. DMIM, RMIM, HD in the same cell.

Don Jorgenson: there are two or three DAMS involved.

Jane Curry: you pull in varous CMETS, giving traceability. You have equivalent thing going. If you expose everything, you get back to wallpaper models.

Don Jorgenson: the design is to have very few paths that drill all the way down.

Jane Curry: this is why V2 makes it hard - they are only explicit when you go to the code. There is a snowstorm in denver, so I am leaving early.

John Koisch: It is in scope - you know that it is not a single service specification, but a series with a sequence of events.

Don Jorgenson: this is a good match to HITSP, so other HITSP can use it.

Charlie Mead: so everything can be layered the same way. HL7 version is it is all in the MIF.

John Koisch: how they surface complexity, early, is part of the specification. it needs to be a shared component. - it usually comes out in the last 30 days. when the project is threatening to shutdown.

Abdul-Malik Shakir: will we have a platform independent specification of the MIF?

John Koisch: the next version of the MIF may be the meta-model for all of HL7. Lloyd thinks it is today.

Wendell Occasio: xml a platform?

Abdul-Malik Shakir: yes. UML is

Wendell Occasio: UML is a language. Using XML in the MIF as wire language as well as expression. The MIF is both.

John Koisch: The MIF is a meta-model for everything.

Wendell Occasio: It cant be for the DAM?

John Koisch: it could be. CTS2 has another meta-model for how vocabulary is expressed on the wire.

Wendell Occasio: you can express XML in XSI.

Abdul-Malik Shakir: I would like to see a UML rendition of the MIF.

John Koisch: Ultimately the MIF should be a platform independent meta-model.

Wendell Occasio: a big meta-model for the SAEAF.

Abdul-Malik Shakir: need an explicit expression of the SAEAF meta-model.

John Koisch: BF?

Abdul-Malik Shakir: only a subset

Wendell Occasio: does not describe the static stuff.

Abdul-Malik Shakir: we have paper power-point/word is the expression of the meta-model.

Wendell Occasio: in parellel we need to formally describe instantion of the meta-model. You are done, it is perfect, how do I give you the answer to all those buckets - UML does not have a way to do that.

John Koisch: It is built for instantion.

Wendell Occasio: an issue with UML is that you cannot easily create an instantiation from it. They made it into sterotypes in CDI?, class diagrams and component diagrams. Otherwise we will end up with a meta-model noone can use.

John Koisch: I have a OWL instantiation.

Abdul-Malik Shakir: OWL, UML, XML are not platforms.

John Koisch: OMG sees XML as platform, but UML is not.

Wendell Occasio: the notation is a notation, not a platform. It is relative, even OMG calls XML a platform.

John Koisch: all models are wrong, all expressions languages suck.

Wendell Occasio: we have to separate expression language from platform.

John Koisch: There is a transform from LDL to UML.

Wendell Occasio: it is not complete. WHat is the point?

John Koisch: Owl is better at some things. but it implies losses.

Galen Mulroney: i have a project to make an effort to bridge the camps of UML and ontology. OWL and protege tools build ontologies, but cant build implementable artifacts. Therefore a bridge

Wendell Occasio: how is the set-membership problem solved? UML creates universe - if I am in Universe M2, the sets are in M1. Automatic transformation?

Galen Mulroney: Research - can we do it, and what does the tooling look like?

Wendell Occasio: answer is in the profiling area.

Galen Mulroney: there is interest among clinician enamored with clinical model - self-describing self-contained models. Open-air models. THey understand business rules about the concept. BP of 5000/-1 is not valid. YOu can do in OCL, but you are switching languages. OWL is better at defining constraints, ADL does it well, but is proprietary. I need to understand the set-relationships, with business rule constraints. It might be impossible

Wendell Occasio: it is in the bounds of the informational viewpoint. I have seen projects lessons learn are that you cannot make it too complicated up front - noone can understand.

Abdul-Malik Shakir: agreement that the meta-model needs to be expressed?

Wendell Occasio: expressed and the language /methodology expressed.

John Koisch: can be human mediated Rigid expression language is automated.

Abdul-Malik Shakir: I think that there will be a need for more than one expression of the meta-model. One expression is targeted at humans to understand. Anothere expression understood by machines. is MIF both?

Galen Mulroney: Lloyd is both

John Koisch: Lloyd is for machines. Grahame says not to worry about the MIF. They think of it as a architecture tools.

Abdul-Malik Shakir: tool builder will need to know the MIF. there are far less tool builders than tool users. I am worried about how to explain the meta-model if I am using XML.

Galen Mulroney: there is no published meta-model. Lloyd says it is, but the MIF is not documented like a meta-model.

Abdul-Malik Shakir: in 72 I thought Cobol was self-documented.

John Koisch: I agree with Lloyd and Woody. Only twice the number in this room care about the meta-model. THe real value of a meta-model, or ontology, is to get our head around it - simple enough for a tool user. COmplex outputs is like haveing no tool at all.

Abdul-Malik Shakir: how do we evaluate the tool if we dont have that.

Charlie Mead: does that sound like how do we certify conformance.

Dale Nelson: the MIF is the message (:-}. Before the MIF there was an access database with rosetree, an expression of the metamodel. The Access database had its own schema which made HMD's possible. We did not worry about it, just the HMD's. Only Woody was worried about the database. MIF has been elevated above the technology - we need to push it back down.

John Koisch: only Lloyd elevated.

Galen Mulroney: the content of the ballot, only the MIF is normative.

Dale Nelson: only the HMD's are normative - table expressions in the ballot. That is how we have elevated MIF.

John Koisch: Jane discribed it as an analysis artifact, not designed for consumption. The MIF is an amoeba that has grown into a blob.

Jane Curry: the MIF is a result of constrained resources.

John Koisch: HL7 is dealing with rich expression and models.

Jane Curry: the crucial nature of the authority is that it is used to determine whether the artifacts are valid to the methodolgy, without having to be educated with a PHD post-doc to do it.

Wendell Occasio: the leve of complexity for the SAEAF cannot be more complex than the RIM, or we will not be understandable.

Jane Curry: the six boxes or the RIM

Wendell Occasio: people understand it, but dont understand the MIF.

Brian Davis: Translating CABIG to what we do - must there be a meta-model of SAEAF? I dont see where all this fits in.

Abdul-Malik Shakir: I am a carpenter that sees everything as a nail - I am a modeler, like to see things in models. I use UML to express designs and analysis. I have two books that contain the UML meta-model. THere is also a schema expression in XMI/XSD. I look at the meta-model, not the XMI. I want to look up the rules, not look at the schema.

John Koisch: can you generate the schema without the meta-model.

Jane Curry: one is machine processable representation of the other. It could be JAVA or schema.

John Koisch: Three guys in a garage dont need models, but a soon as others are involved - we have 20 neighborhoods.

Wendell Occasio: is there value in standardizing the nature of the artifacts that fit the cells? Precisely.

John Koisch: for almost every cell. Long term, everything.

Wendell Occasio: that is the meta-model.

John Koisch: a process in Oracle/Microsoft a huge effort for software to write software. The process ended from the internal people as a failure, since they did not have a meta-model. We are trying to get to an expression that can allow everything to be automated.

Tony Julian: automation is only possible if we constrain it - currently I can use an HL7 schema, shove it into xmlspy, generate CJ++ classes, which will feed visual studio to create code. BUt the number of classes is very large.

Abdul-Malik Shakir: we have agreed: 1.we need to express meta-model. 2. expressed in human and machine 3. MIF does machine well. Is it worthwile to create a Human expression model worth the cost.

Galen Mulroney: i yeild to Jane.

Jane Curry: my point is the reason there was no attempt to maintain concurrency past the first HDF in human readable form, at the same time Woody and Lloyd were working on an independent tools set trying to make it compatable. The toolsmiths went one direction, humans went to lore, so the MIF manages the complexity in our process and artifacts. When no volunteer came forth to upgrade UML to UML 2.1, the work dropped off. SO we have an obsolete UML profile. It is XML representation of UML instances that allow us to to produce our stuff. In some places the conventions have been broken.

John Koisch: the modeling languages are superior to UML - things UML cant do.

Jane Curry: UML does a poor job of vocabularies.

Galen Mulroney: I have seen SAEAF conformant. WHat is it? THere must be a meta-model by which we can define the conformance.

Wendell Occasio: you need a meta-model to standardize the SAEAF. That process happened ground-up.

Jane Curry: there are candidate artifacts in XML

Wendell Occasio: it was a way to express what was happening.

Jane Curry: it was supposed to go on paralell with the HDF effort.

Wendell Occasio: it was coding and implementation

Jane Curry: there is a bunch that uses the principle of designing from smaller grained.

John Koisch: the way they built it was from bottom-up. THey were concencious.

Jane Curry: using XML for expressions.

Wendell Occasio: did HMD and constraints exist before the MIF. FOr SAEAF, we are introducing new concepts - we have an additional burden to communicate the concepts, explain the usefullness. Opinion - express of the endeavor is inversely proportional to the meta-model.

John Koisch: I disagree. The MIF exposed a lot of things. It has the aspects of a prototype project - exposed how hard conformance was. The MIF was used as a machine executable explicit expression. The SAEAF is simple, two things - conformance and compliance, and a governance model.

Wendell Occasio: not it is not.

John Koisch: the reason for the BF is that SOA had taken it on, but MnM said to come up with dynamic model for SOA. John Quinn said to come up with the behaviorl model. The real point of the BF is the reason the MIF failed is that it implied things about the computational viewpoint.

Jane Curry: it made few rules which it inforced.

John Koisch: In summary, the fact is if we had had a BF, we would not have had to do so. The BF is extension of the conformance model.

Wendell Occasio: it is complex, it needs to be simple to be adopted. MIF is too complex. I agree with machine computability, but it is icing on the cake. People willnot adopt just because Arb/TSC says so. It needs enough precision so it can be tested manually. Then to be really successful it needs to be machine processable.

John Koisch: the real problem is not that it is hard - 20 years ago they said it was easy - now after spending time and money, we still cannot interoperate. We can just talk about what it means. The essential hardness cannot be reduced - after you undertake the hardness, maybe there is tooling, template repositories, even after you have done all that, regardless of the developers - cross functional teams, the bad thing is when you come to the point of "What did i do that for?". THe architecture proves the cost/benefit of the effort.

Wendell Occasio: It is too complicated.

Abdul-Malik Shakir: i like your list. THe meta-model needs to be expressed to find the rules, the rules need to be explicit enough. We will need to manage the evolution. MnM used the meta-model to understand the rules. Once it was machine processable, we relaxed the processes: we have to tell the people what the tool already knows. DO we need to add compatibility? lets look at the model. We need a blueprint.

Jane Curry: we know complexity lives in the simple framework, which can be simply described. Contents of the cells, and viewpoints, is not as simple as we would like it to be. It should be as easy to understand as the RIM is. You can describe it in 10 minutes - the pattern is replicated in everything else - it is a simple pattern. If our framework is that simple, we should be able to describe the metamodel simply, hiding the complexity. The core reason we are doing this, what information with whom to achieve what purpose. That is the underlying thing. All of our framework is oriented to expressing those things at different levels of specificity to target different audiences.

Wendell Occasio: we are trying to solve the often-failed dynamic model. Right now, it is hard to adopt/swallow.

Abdul-Malik Shakir: the meta-model helps us to find holes. We found it useful to have a UML model (can we have recursions, how milestones relate to collaberations). I noticed an association between milestone and activity. Without graphics you cannot see the holes?

Wendell Occasio: i am trying to cut/cut/cut. Milestone is an attribut of interaction. The attributes of milestone may be pushed to interaction type. Is the cost of the extra class justified.

Jane Curry: I thought you wanted a simple meta-model for SAEAF. THe crux of the model is the BF.

Wendell Occasio: the crux is in the BF: less about creating the artifacts.

John Koisch: I dont know how much I need to see this.

Wendell Occasio: If you dont understand the RIM, you cannot build messages.

John Koisch: we need a simplfied version for the committees. Trigger-type initiates an activity relationship. Is the trigger type valid for collaboration, interaction, and exchange? In terms of best practices, it should be tied to interaction. You need a generalized schema/expression language, and a specilized model. Everyon does not need the richntess. Trigger-type/activityRelatioship is a super-type.

Jane Curry: the triggers are starting to act at an organization level. There are impled collaborations that become explicit.

John Koisch: clinical practice becomes the point of granularity.

Jane Curry: if it is not a the right level, it is disruptive to change later. It is like you are specifying a diagram. We need a simplified version - a schema that needs to be filled out, and here are the triggers.

Jane Curry: it is like role-ing. the best practice is dont use role-ing if you dont have to.

John Koisch: never start a sentence with a preposition, but sometimes it is right?

Brian Davis: for CABIG dont you need a very simple thing with a diagram and a lot of words. This would go to modelers, no matter how many time your describe the rim, people who dont use it will forget.

Abdul-Malik Shakir: you extended #2 it needs to be human-target expression(S). one for each audience.

Wendell Occasio: making it simple is hard.

John Koisch: if a college freshman cant understand it, it is two hard. Do we still need a 400 page document?

Abdul-Malik Shakir: yes for an audience

Wendell Occasio: the metamodel needs to be simpler.

John Koisch: Charlies priority is getting the 400 page document.

Wendell Occasio: priority is the decks - to evaluate.

John Koisch: he is telling me and Jane we need to write the documents.

Wendell Occasio: lets get the decks and models to doable.

John Koisch: some of the decks are re-usable. Out of sync with the documents.

Abdul-Malik Shakir: goal is to create the document. Not everyone can read it.

John Koisch: the 400 page document is in this room, and MnM.

Abdul-Malik Shakir: I need the 20 slide deck.

John Koisch: you are saying something different.

Abdul-Malik Shakir: we need to do the models first, otherwise the text will have errors. I need the text/description, with examples and constraints. I can create a document from the tool, that describes links and content.

John Koisch: at NCI we are publishing our models.

Abdul-Malik Shakir: good modeling is good, publishing is better. I am concerned about the notation.

John Koisch: it is a standard in modeling error handling - you stereotype the throws. I borrowed it.

Abdul-Malik Shakir: why do some associations have navigability, in some places you cannot. Can you have an milestone without an interaction type.

John Koisch: all interactions have a relationship to a milestone

Abdul-Malik Shakir: every interaction has a milestone.

Wendell Occasio: we can simply.

John Koisch: do we agree that this is universal, then we can constrain it.

Wendell Occasio: that comes with a price.

John Koisch: you have a series of related models.

Galen Mulroney: VA is generating different models for different audiences. We could have this for the ARB/SAEAF only, another where half the boxes and sterotypes are gone. It has been successful. Early uber-model was confusing everyone. So if we can generate through tooling the other models.

Abdul-Malik Shakir: Move that the ArB commit to using an UML model to represent the architecture. Second Jane. I dont know what will be written, but it just does not look right.

Abdul-Malik Shakir: Only milestones have names and types.

Galen Mulroney: if you wanted to suck milestone up, you could rename.

Jane Curry: we need to start with a simple concept diagram, and their inherent relationships, from which we can drill down.

Abdul-Malik Shakir: we have removed work-unit? It was there yesterday.

Wendell Occasio: i agreed, but I am surprise to see it.

John Koisch: work unit is gone. Wendell stated that work units are collaborations. The point of WU was a CMET of behavioral stuff, then call it collaboration. It is redundant. As a developer I will use work units.

John Koisch: call the question?


Abdul-Malik Shakir: Motion: We need to define a controlled process by which all of our the model is changed. Tony Seconded.

John Koisch: It is under software change management. It has been published many times.

Abdul-Malik Shakir: the change process makes us aware of the change.

Wendell Occasio: that assumes a baseline.

Abdul-Malik Shakir: you are describing a control process. That has to be defined. Who can propose changes, who approves/denies.\

Wendell Occasio: how is that different from the slides.

Abdul-Malik Shakir: a while ago we said Galen cannot vote on it, can he suggest the change? I dont know what the rules are.

Wendell Occasio: my assumption is that individuals have been tasked, to modify things that have not been finalized.

Abdul-Malik Shakir: it is still in development.

John Koisch: out of this meeting all of this will be released.

Wendell Occasio: we are not ready for that baseline.

John Koisch: there is no reason you cant be informed when changes are made.

Abdul-Malik Shakir: i would like to see a process like the RIM. Eventually.

Jane Curry: if this has the same standing, then changes will have to have the same rigor. We have not looked at it. We are challenged to review it in time. I agree that the changes need to go forward, and the rational be transparent. If Wendell made a comment on Version 3, the model picture has not changed since February.

John Koisch: BF started in July, worked until august, conference calls in October, Changes in november were published, in december Abdul-Malik, John, Allen Honey, and two others made changes, in february there was a publish based on changes in the HL7 meeting

Abdul-Malik Shakir: it is your task/priviledge. If we are going to adopt it, we need a formal change process.

John Koisch: you cannot walk into the room and decide it needs to be changed.

Wendell Occasio: in development you need to change.

John Koisch: ammended to:

Abdul-Malik Shakir: Motion: We need to define a controlled process by which all of our core models are changed. Tony Seconded.


Abdul-Malik Shakir: at this point, they are a work-in-progress.

Wendell Occasio: ammend to once it is an official document.

Abdul-Malik Shakir: Motion: We need to define a controlled process by which all of our core models are changed once they are official ArB documents. Tony Seconded.

 Abdul-Malik Shakir: Motion: We authorize Tony to propose a draft controlled process by which all of our core models and documents are changed once they are official ArB documents.  Tony Seconded.  QC includes measure points, and what does 'official QrB documents mean?
 Motion by Tony to call the question, seconded by Craig. 8-0-1
 Motion called 8-0-1


John Koisch: slide decks are authoritative. It has been messages.

Jane Curry: priority of things to be done. Wendell and John were discussing the problems with the behavioral model.

John Quin: John/Charlie need to report to TSC on monday. There is effort pick up pieces of this and use it. the Logical Record Architecture.

Jane Curry: the BF is of interest to Canada Health Infoway.

John Quin: we have alpha users who are primary stakeholdes in the organization.

Jane Curry: we said "in the end, this is what we need - a simple model".

Abdul-Malik Shakir: is not this th 4x4 matrix?

Jane Curry: 4x4 is a thumbnail of our methodology. Jane read four bullets:

John Quin: data, process, reference terminology, and transport.

Jane Curry: as a picture, we can say this is how we drill down.

Jane Curry: reference terminology is horizontal related to all the others.

Simple-arb-model.JPG


Abdul-Malik Shakir: Unclear about now and Kyoto. Are we done? What needs to be done by whom?

John Quin: progress has been made, priority expressed. Relationship from HDF to SAEAF is cloudy, as is tooling and core-principles.

Jane Curry: Big hole is in BF. implication of this for the alpha projects are for how the alpha projects deliver when our publishing process is not adequate. So how do we recommend they do it? We validated the discovery that either we add more slides and examples, or create placeholders. The artifacts are higher priority in the slide deck, rather than the paper - required by the 23rd.

John Quin: the huge slide decks are only for this room - they are fine, but i dont want expectation that without a year education on the SAEAF the slides are sufficient.

Galen Mulroney: the papers are important.

John Quin: papers for ballot, pictures are for this room.

Abdul-Malik Shakir: we created outline of the kinds of things to describe in the cells.

Jane Curry: conceptual/computational cell contains the outline we developed.

Abdul-Malik Shakir: do we still know what needs to be done in this cell, how much is done, and to be done?

Jane Curry: i have draft not reflected, so i am expecting to include those.

Brian Davis: outline of suggestions for each cell's contains

Jane Curry: will not have all filled in by Kyoto. Conformance statements should be declarative statements with Yes/no answers.

Brian Davis: I am going to - there is a project in CABIG - which created artifacts and conformance for each of the cells. and fill in the concrete examples.

John Quin: we dont have anything on practical instantiation.

Brian Davis: John, paul, Charlie and I will be working on it.

Jane Curry: if you assemble into a document, we can peruse them as we work on our stuff - send to john or charlie.

Jane Curry: the cells, and the draft documents - dont think I can have by Kyoto, just decks and cells.

Abdul-Malik Shakir: we have three thursdays between now and Kyoto- so we will not have a call on the seventh, so we only have two calls. On the 30th we should know where we are at.

Jane Curry: Posted on the 29th will be what you get.

Abdul-Malik Shakir: planned activities: TSC on saturday?

John Quin: first two quarters of saturday are OCM with Mark.

Abdul-Malik Shakir: expect SD to expect something. Are there others to talk to their division? What are we planning?

Galen Mulroney: I will be there. I understand there will be an all-hands meeting on Wednesday, one quarter, subject SAEAF. Includes all of the members of all the SD's.

Tony Julian: Inadequate communication of the meeting. Check with HQ, John, and Charlie on how to do. Tony Julian