Hi Marcus/ Pablo,
I think the comparison/ contrast to Github is instructive, because, of course GitHub is a hugely successful product which is highly supportive of open-source development, but it is not itself open-source. It is a proprietary tool. If you truly feel that tooling to support collaborative working itself necessitates an open source license then you should close your Github accounts and look elsewhere.
I would very much like to see a future where levels of sponsorship, industry engagement, national funding etc, etc made it possible for CKM and other similar tools to be open-sourced but we are simply not in that position right now. All of the key authoring tools are open-sourced or free, (and all, I understand, will be open-sourced within a short period).
CKM was built to perform a very specific role i.e to help informaticians manage the complex process of crowd-sourcing clinical input, working out the impact of version changes, handling translation work, term-binding work, terminology building, particularly at international or national level. It is not needed to build an archetype, build a template or build a termset. It is not needed to display an archetype or template or termset. All of the resources are mirrored to GitHub and all of the specifications and information necessary to perform these activities are freely available.
CKM is a highly specialised tool with limited focus, primarily on national and international asset management. It is not needed to build openEHR systems, any more than GitHub is needed to build open source software.
Alternative repository management tools are starting to appear, such as the 13606 Assocn. CIMM. I am sure David and Diego will not mind me saying that, as things stand, CIMM is a fair way off providing CKM -style functionality.
I think we are in danger of confusing some real and significant issues around community engagement with the Foundation governance process. The issue of CKM licensing is model has, in my view, no practical impact on the concern that Pablo raised. Don’t confuse the tool with the process.
Even then I think we need to be aware that there are probably two quite different requirements here.
We need a much better way for good candidates for international archetypes to find their way into the international repository, probably to Incubators in the first instance. Some of the upcoming technical changes to the tool will help this but we also need to develop clear policies of how and when this is appropriate. The Foundation repository is primarily designed to manage set of archetypes as a ‘source of truth’ with new content flowing through in a relatively controlled but coherent fashion. Managing the governance of these ‘semantic assets’ requires much more care and precision than ‘source code’
This is quite different from the position in e.g Github which is essentially a tool which allows some degree of socialisation between otherwise siloed repositories. This is great for allowing assets and source code to be exposed, forked and re-used but it lacks the control and coherence that is required by ‘managed’ national and international standards development.
I actually think we need both kinds of environment, and there is nothing to say that both environments need to be instantiated in the same tool.
@Marcus - there is actually very little metadata in archetypes. The translation support that Silje asked for is already supported in the AOM, and in some archetype editors such as LinkEHR. It is not supported in the openEHR archetype editor but as this is an open source tool, I will be working on that problem later today :).
I think there is a lot to be said for using Git to manage some of the versioning and asset management activities we need, indeed I do that all the time when working on local projects, but none of this kind of metadata is carried in archetypes anyway. The kind of versioning and governance metadata that we do need is equivalent to the metadata used by RubyGems or npm, needed for distributed source control, and the new versioning metadata that will be carried in archetypes is compliant with Semver which underpins npm.
ADL is actually a very readable language, given the complexity of information it needs to convey.
It is, of course, unfamiliar but it is perfectly possible to produce xml, json, yaml … serialisations of the Archetype Object Model which is the real source of truth.
XML serialisation is fully supported by the LinkEHR and openEHR Archetype Editors, Thomas’s Archetype Workbench exports these other formats and the template designer output is all expressed as XML.
The problem is that these non-ADL serialisations are actually much more difficult to read and understand than raw ADL, once, of course, you get your head around ADL.
@Pablo - CKM does make use of a proprietary document management system but the real challenge here is not technical, it is how we find a funding model that would sustain the kind of professional support that a tool like CKM requires. This is not a hacker project, it requires sustained investment, proper maintenance and a proper business model. So far it has not been possible to persuade the wider informatics community to collaborate on the kind of joint funding that would make commercial sense to a prospective supplier.
This is an important discussion. I’m glad to hear people being supportive of all the great work that has been done, particularly by Heather Leslie and Sebastian Garde. It is not easy to develop a first-of-kind product.
I think we have a great opportunity to discuss how to expand CKM editorial capacity, review current editorial policy around community involvement and to see how other non-CKM applications might fill some of the gaps that have been identified. I will certainly raise this via the new Board and, of course, discuss further with Heather in our capacities as CKM editors and Heather’s position as Clinical program lead.
Let’s not mix that discussion up with an equally important issue of how we can secure the funding necessary to sustain development and support for repository tooling in the future. I don’t think there would be much objection to the principle that an open-source licensing model would be preferred but that can only happen if the commercial model makes sense for potential providers.
Ian
