The concepts of collaborative development aided by revision control systems have been well known in the software industry for decades. Some of these systems have been adopted in terminology authoring tools like the IHTSDO Workbench to support collaborative authoring workflows. This talk discusses implementation challenges and lessons learned when applying collaborative development techniques to terminology authoring. Representative terminology authoring use cases from the Singaporean, Australian, and Canadian National Release Centers will be discussed in terms of their impact on revisioning requirements. A concrete scenario for SNOMED CT will be demonstrated including examples for versioning, comparing versions and patching an older version. The scenario will be driven through a representative collaborative workflow. Useful open-source components will be discussed along with practical experiences in integrating them with Snow Owl, a commercial terminology authoring application. Revision control features including change history, comparison, versioning and patching of the terminologies will be discussed and compared to alternate approaches. The impact of workflow to drive the process will be illustrated using the tool. Optimizations and performance challenges will also be briefly covered. Please see our website http://b2i.sg for further information.

1. Revision-controlled collaborative
terminology authoring
IHTSDO Showcase
2014.10.30 - Amsterdam
Mark Czotter
Balazs Banfai, Ph.D.

2. Our approach
• Real-time collaborative authoring support
– Not an afterthought, the repository IS the store
• Workflow support
– Externalized therefore replaceable
• Cover the entire domain of healthcare terminology development
– All terminologies and all terminology components
• Including local terminologies, value domains and mapping sets
– Information models and terminology bindings
• Modular and extensible
– Capabilities are readily available for extensions
• Multiple deployment modes
– Stand-alone, thick client-server, thin client-server via REST

3. Collaborative authoring
• Multiple authors can work simultaneously on their
tasks in their dedicated and isolated ‘sandboxes’
• Completed work can be promoted to main repository
• Sandboxes can be synchronized to reflect relevant
changes in the main repository
• Potential conflicts are handled
• Authors receive notification about relevant changes
immediately

4. Workflow support
• External, managed by an issue tracking system
– Loosely integrated with ‘hooks’
– Bugzilla, JIRA, etc.
• Terminology authoring is only a part of the
overall enterprise workflow

5. Revision control features
• Terminology repository
– Each terminology has a dedicated repository
– Each commit becomes a revision
• Revision history is maintained
• Changes between two revisions can be compared
• Commits can be reverted
– Supports branching (patch), merging, versioning

6. Revisions on a branch
MAIN branch
Each commit is
stored as a revision
The last commit
point on the branch
is referred to as head
A CB D

7. Tasks
MAIN branch
Work associated
with a task is always
committed to a
dedicated branch
Task branch
A B
B2
B1

8. Isolation
MAIN branch
A C
Task branch
Terminologists working on the MAIN
(head) can see changes by: A, B and C
but cannot see B1 and B2.
B
B2
B1
Terminologists working on Task 1
(head) can see changes introduced
by commits: B2, B1, B and A but
cannot see C.

9. Promoting
MAIN branch
If deemed worthy,
content of the task
branch can be
promoted to MAIN
Task 1

10. Synchronizing tasks with MAIN
MAIN branch
A C
Task 1Commit C can be relevant to the work on
Task 1, the content on Task 1 should be
synchronized with MAIN.
B
B2
B1
After synchronization commit C is
visible on Task 1 and the content on
Task 1 can be promoted to MAIN.
D

11. Typical scenario
MAIN branch
A C
Task 1
B
B2
B1
Task 2
C1
E
Task 3
D
C1.1

12. Versions
MAIN branch
Task ‘Patch’
Version branch
v1

13. Task
Management
Content
authoring
Task created for
1A&1R scenario
Review
Authoring completed
Task is set to Resolved
Admin
Author
Reviewer
Reviewer accepted changes
Task is set to Verified
Reviewer rejected changes
Task is set to Reopened
Single author and single reviewer
Content
promotion
Content is promoted to
MAIN repository
Task is set to Closed

14. Content
promotion
Task
Management
Content
authoring
Tasks are created for
2A&2R DIA scenario
Final
Review
Admin
Author
Reviewer
Final review completed
Task is set to Verified
Dual author and dual reviewers
Dual independent authoring
Content is promoted
to MAIN repository
Task is set to Closed
Content
authoring
Adjudicator
Review Review
Reviews completed
Final review completed
Tasks are set to Reopened
Authoring
completed
Tasks are set to
Resolved
Changes
rejected
Task is set to
Reopened
Changes
rejected
Task is set to
Reopened

15. Demo
• Create Task (Single author, single reviewer authoring)
• Create new child Concept under Tetralogy of Fallot
– Change to MAIN to show that concept is not there
– Change to Task to show that concept is there
• Review Concept
• Show history view
• Show commit info

16. Demo cont.
• Version content
– version: Showcase_Version
• Compare with previous SNOMED CT version

17. Questions?

18. Revision control

19. Workflow

20. Versioning and Compare

21. Revision control technologies
SVN Git CDO
+ Checkout subtrees + Small and Fast + Modeling, less
development
+ Locking + Distributed + Branching
- Branching is hard + Branching is easy + Locking
- Text based - Text based - Migration is hard

22. Task N Task 1
National
Extension -
20140131
Complex branchingInternational
20140131
20140731
National
Extension -
20140731
Migration
NRCs use
version to
create their
extension
NRCs migrate
their extension
when new
version is out