On 14/12/2012 00:29, Bert Verhees wrote: > > You make the system depending on the quality of organizations. I really did > explain it several times. Please read my latest reply to Eric Sundval, > Tuesday, it was, I think. It must be in the archive.
I think there is no other alternative, at least for now. > My problem is that the OpenEHR datasets depend on archetypes and they express > this dependency by the property Archetype-ID in the Locatable-class. >> yes, that's the current spec. As I said elsewhere, there is a proposal (for >> some years now) to improve this, and to enable the use of the full >> multi-axial id (generated) and also Uids or Oids or whatever. So I don't >> think that in the future, there will be any problem. You will of course have >> difficulty to see what the archetypes are when inspecting data, so that's >> the downside. > OK, this is the good news I was hoping for. > > Do you know why this proposal never is discussed in the mailinglists and why > it takes years to implement it in the specs? well I think it has taken time for vendors (including you, and others!) to get implementations together and working, to get a better feel for what the right approach is. I could generate a theoretical proposal (well, in fact I did ;-) but I am happy that we never implemented it directly; the last 3 years' experience across real implementations has generated knowledge about how to proceed properly. Now I think a proposal that we all put together will really work - we can do this work pretty quickly. > > It is not only OpenEhr that is suffering from this weakness, also En13606 is. > Why isn't this brought under attention while it was under ISO, 2008, or since > then? well ISO just isn't connected to implementers in any meaningful way. That's an argument for another day, but essentially the whole model of how standards are done in ISO needs to change, at least IT standards. Stephen Kay has ideas on this. I think the best we can do with respect to ISO today is to provide quality proposals from openEHR for the revision of 13606 - stuff that works, like a better EHR Extract etc. > > It is a weakness which can make datasets useless, especially in a > message-structure, when data leave their safe eco-system. I wouldn't say 'useless', but you are right - there are risks that should not be there. So just like all the official standards, then ;-) We need to do better of course, and starting in 2013, we can start work on the specs again, and it really will require the implementers to step in and contribute ideas. > > The key feature, semantic interoperability is at risc, because, when > receiving a data-set, how can you be sure if it does refer to the same > archetype you have in your house? There is no way, no digital signing, no > checksum, no centrally enforced registry, nothing. > Even asking the archetype from the organization which sent the dataset, which > is cumbersome, it still is no guarantee. various implementers have already hit this problem and solved it in local ways - which is a bad outcome. We need to rectify that. We are nearly set up for doing more work on the specs and software - it's taking time to do some practical things - new website, move everything to Github, and define new governance. I share the frustration, but we are getting there. > > And that, while there are more solutions possible to resolve this. > > I think this is critical. Don't you? > > I did a proposal in the same email, I referred to, to Eric, above. Only an > idea to handle this. You can easily find it in the mailinglists-archive. I am > sorry that I don't have it at hand at this moment. > > But of course there are more ways, and maybe better ways, to solve this. just make sure you are ready to contribute these idea when we start working on the proposals! - thomas
