Reference Implementation Meeting 2010-07-06

Notes from Reference Implementation Workgroup
Date: July 6, 2010
Time: 12pm-1pm
Attendees: George Cole, Didi Davis, John Moehrke, Chris Moyer, William Lusen, Doug Arnold, Don Jorgenson, Arien Malec, Uvinie Hettiaratchy
Actions for this Week

#	Date	Action	Status	Owner	Due Date
2	07/06/10	Draft initial open source policy for reference implementation work	Open	Brian	07/06/10
3	07/06/10	Create a matrix/list of components to prioritize and determine scope for reference implementation work	Open	Arien	07/13/10

Agenda

• Kick Off Meeting

Notes

Arien Malec
· Main purpose of the group is to oversee work of creating strong reference implementation or set of implementations for the consensus proposal.
· Specification work not yet done, which might be helpful. Writing a REST specification while doing a REST implementation was helpful.
· I want everyone to recognize that we’re doing reference implementation in parallel.
· One of the first things we’d like to decide are the components of the reference implementation.
o Do we have too many reference implementations we’re taking on?
o Microsoft has a .NET reference implementation
o Many organizations have Java implementation
· Want these in parallel or one? Let people be free to create their own?
· Second key decision: What are they key components of reference implementation?
o Need to decide overall priority
o Which components are in scope and not?
o Whose doing which pieces of work?
· I’d like to pose these to the group
Brian Behlendorf

• This is where the rubber hits the road in this project.
• It’s really important to have a clear sense of where we want to go and if we have resources to get there.
• The conversation about multiple implementations and the environments on which we build them on is important.
• I want to focus as much as possible on laying out a pilot technology plan and doing some advocacy within our organization for resources in addition to ourselves, maybe look outside of our organizations as well.
• I would like to suggest that we focus on the pilot technology plan – the immediate follow up on the consensus proposal
  • Talk about components we’re committing to
  • Having approachable end user interfaces

Round the Room: Feedback/Comments on Scope of Reference Implementation

Name	Feedback/Comment
George Cole	Like idea of focus on priorities. What I don’t know is when the pilots need to kick off. I’d like to ask to add the complete calendar from front to end. Arien Malec – We need to revise the calendar.
Didi Davis	I like idea of scoping. It will help to understand the calendar. From our perspective, we may have some partners but not sure how they will add value yet.
John Moehrke	Nothing further to add.
Chris Moyer	As we talk about components and priorities – one of the biggest challenges will be architect framework solution as part of reference implementation.
William Lusen	Question on the environment where we’ll implement the reference implementation. Arien Malec - Critical scope issue. I would suggest goal of reference implementation is to be a library of implementers to allow details for testing infrastructure. Organizations may take that reference implementation and use it as an introduction. Not a production system. It may serve as open source quality Brian Behlendorf - I think the question was about the pilot and what’s being assumed about where it’s located. Is that the question? Yes, whose going to provide the platform through which we’ll run this reference implementation? Brian Behlendorf - Short term, the focus should be on making the pilot successful – meeting them where they are and what the resources are to carry them to implementation. Long term, what we can do to maximize the impact of this and make it easy for it to be incorporated elsewhere. There will be tension between those two goals, but if there’s a decision point, I would suggest we bias towards the short term. That might mean going to implementation geographies and standardizing on single or small group of platforms.
Doug Arnold	I’m on the implementation geographies group and I’m concerned that as we move forward with our work, we’re going to move forward without coordination. We don’t want to be out of phase. Arien Malec - Good point. I think the implementation group is where this comes together. A number of us also participate on a number of calls and we can help navigate.
Don Jorgenson	I’m concerned that this differs from original platform. Where’s the touch point where we come together?

Arien Malec

• Why don’t we go to that topic. I think Brian raised this. We want to make the right decision with regards to platform or platforms that will engage largest number of development community.
• I think there are two choices
  • a) a Java based reference implementation and
  • b) parallel development of .NET and Java reference implementation.
• I think Option A may get a larger pool on one platform, but we will be leaving off a significant number organizations that develop on a .NET platform that do not have the ability to develop on Java.
• I’d like to elicit takes on pros and cons.

Brian Behlendorf

• I think that’s a right way to frame it. I think we can support these. We know what we need to build. I think it’d be possible within this project to do that. My preference would be to commit to both.

Round the Room: Feedback on Parallel Development of .NET and Java reference implementation

Name	Feedback/Comment
George Cole	I think if we begin to focuse on interfaces first, we can find that parallel implementations of Java and dot net doesn’t mean we have twice as much work. I think we should target that particular choice as library deliverable. Vote for B
Didi Davis	I think we have to look at both to allow for enough connectivity for edge systems. On behalf of state of TN and Carespark, based on vendor responses, we’re going to have to allow for both. I like the idea of focusing on the interface aspect or the edge system pieces.
John Moehrke	Vote for B. Want clarification if there may be places where it’s more obvious that Java is going to be what people want in the HISP. We’re not necessarily saying that everything we build has to be done in both implementations. There may be overlap of some that will be preferential for one or the other. Arien Malec – I do know of at least three organizations that I know would put in the HISP box for a .NET platform. Not going to be that easy.
Chris Moyer	I think we’re behind Option B. Lot of value of adding Java implementation side. Having two people tackling problems on different platforms with different approaches and allowing them to communicate will allow a lot of improvement in quality of our work. Even though as a company we will be able to participate more in Java, value in both. *
William Lusen	Option B as well.
Doug Arnold	I support B. I’m also concerned about making sure we communicate with HIEs which are going down one track and setting up relationships with vendors. Want to make sure ONC is also aware of these. Arien Malec – We should work with State HIOs so they are aware of decisions that we’re making. My comment is general is that there seems to be a lack of understanding of NHIN Direct.
Don Jorgenson	Support B and note that there’s probably not all that much to share between .NET and Java.

Arien Malec

• I would endorse, as Chris said, that doing parallel implementations do allow for more clarity.
• So we decided we’re going to take on parallel implementations on a .NET platform and Java platform.
• We then need to turn to components in scope. I’m not sure we can do all of that on this call. I would suggest that someone create a matrix/list of components we can then prioritize and scope in and out on a future call. I will volunteer.
• My feeling is that the highest priority component would be one that would allow us to go to end to end SMTP backbone testing. The highest subcomponent of that is the security agent. That’s a critical piece of infrastructure for a number of reasons – this is the lynchpin of the system from a security perspective and needs to be reviewed early.
• If we get that component right, it makes the rest easier. The security agent implementation agent should be the highest priority work. I’d be willing to look for feedback or counter suggestions. I realize that’s somewhat contrary to the prioritization step, but my guess would be that would be number one.

Brian Behlendorf

• It’s likely this is where interoperability issues will be most complex. Getting this implemented in .NET and Java early would be helpful. Do we anticipate this group being the one that delivers the technology solution?

Arien Malec

• I will ask the implementation geographies group. Part of the work is creating the matrix of whose supplying the HISP. We can just ask if they’re implementing on this on their own or contribute to reference implementation. Those organizations will help us answer that question.

Brian Behlendorf

• Absolutely. I’d like to make sure that the people who are signing up for that role in the implementation geographies are on this call.

Round the Room: Feedback on Security Agent as first priority

Name	Feedback/Comment
George Cole	Pass
Didi Davis	I agree with you Arien. We have some resources here that might be helpful. Please let me know if we can help. I think the fleshing out part we need to add is that the Tiger Team needs to be aware and cross-communication happens with policy issues. Arien Malec - I agree, I actually talked to them this morning and that should be happening.
John Moehrke	I think the security agent is essential. I just want to caution that there’s not only one way of doing something. I’m not sure what other components we can add on as needed.
Chris Moyer	I have a question since I have not done work on XDR backbone. Was the security agent a portion of this? Did this have same mechanism as DNS? Arien Malec – Content encryption was not part of concrete implementation. If it was, it would use a different mechanism – WBS security, etc. John would know better. I do think security agent is the highest priority. *
William Lusen	Agree this makes sense. I like idea of coming up with a brainstorm list of what we don’t know in parallel. A concrete list might allow us to know what might come next after highest priorities.
Doug Arnold	I’d like to raise a global issue – we expect in next week the final meaningful use regulations will come out. I wonder if we’re looking at scenario of NHIN Direct use cases, especially on ones that concern patient, to have any implications for scenario B for communications to and from patients? Arien Malec - With regard to tech decisions we’re making, I believe they’re non material to user stories we created. By non-material they don’t lock us in or out of any user stories. These were particularly chosen to do a good map to relevant meaningful use criteria. My comments are uninformed. I would not expect any of the final regulations to add or materially change any of the previously posted material. I would suggest a subtraction or optionality. Reasonably good shape with user stories.
Don Jorgenson	I think we can address this early. Arien Malec - We do have a team right now working on specifications for the security agent which is another reason to start. Who and how will this be done? Arien Malec - The two parallel teams who did original security agent were drafting set of specifications prior to holiday. I’m expecting in documentation workgroup meeting, we’ll get a good read as to where that work is. That’s something we’re definitely interested in.

Arien Malec

• A bit of housekeeping, we’ve made two decisions:
  • 1) Doing parallel implementations on a Java platform and .NET platform.
  • 2) We’re going to concentrate on security agent. Being clear if that’s the only interoperable mechanism we’re supporting and being clear about final specifications that we create.
• The final bit of housekeeping is to clarify the source code license. The concrete implementation was under a BSD license?

Brian Behlendorf

• I think that should be license we can pursue this under. Is it ok to have a reference implementation under a proprietary code, especially the .NET one? I could do a draft of a policy that might work.

Arien Malec

• That would be great. I propose as a decision that reference implementation will be released under a BSD license. Brian will draft initial policy.

Brian Behlendorf

• I will do a draft of that by today.

Round the Room: Brian to draft initial open source policy

Name	Feedback/Comment
George Cole	Agree
Didi Davis	Agree
John Moehrke	Agree
Chris Moyer	Agree
William Lusen	Agree
Doug Arnold	Agree
Don Jorgenson	Agree

Arien Malec

• I think we have a good set of actions to work on. Any other questions?

John Moehrke

• Could you disseminate to all the workgroups the names of all the organizations that have agreed to be HISPs and if they’re geographic or national?

Arien Malec

• I will ask the implementation geographies workgroup to do that.