Stream: IG creation
Topic: IG Publisher Versioning
Grahame Grieve (Feb 01 2022 at 21:53):
The IG publisher is at 1.1.98 currently. It's kind of a bit hard to version the IG publisher because there's no big ambition for structural changes to it, just a never-ending stream of small features and bug fixes. So I'm really not sure when it would be appropriate to increment minor or major versions.
And there's no reason for binary compatibility for the publisher API to be an issue either, so @James Agnew has pointed out that semver isn't necessarily appropriate. Given that, I'm wondering whether we should version the IG publisher the way smileCDR is versioned:
YYYY.MM.N
Where n serially increments and resets to 0 when the YYYY.MM changes
Grahame Grieve (Feb 01 2022 at 21:53):
opinions are welcome
David Pyke (Feb 01 2022 at 21:55):
I think that works but I'm not sure that how the version of the publisher is represented is terribly important, as long as people can get the latest version easily.
Grahame Grieve (Feb 01 2022 at 21:56):
that's why I haven't got around to doing anything about it
John Moehrke (Feb 01 2022 at 21:56):
so your goal is to just move to YYY.MM.N to give a reason to not just forever incrementing the last digit?
Grahame Grieve (Feb 01 2022 at 21:56):
I could just arbitrarily increment it to 1.2 once we get to version 99, like I did for 1.0 to 1.1, but I thought I'd get around to thinking about it
John Moehrke (Feb 01 2022 at 21:57):
works for me.... but what happens when the year 9999 is done?
David Pyke (Feb 01 2022 at 21:58):
Then we'll be on FHIR R928 and the IG publisher will be a psychic interface and we'll just know the version
Grahame Grieve (Feb 01 2022 at 21:58):
I'll be surprised if that's a problem for me
David Pyke (Feb 01 2022 at 21:58):
I like that you'll be surprised in the year 9999. That's long term planning
John Moehrke (Feb 01 2022 at 22:00):
well, look... he did ask us for our opinion like we were going to give him useful input.
Chris Moesel (Feb 01 2022 at 22:21):
Wait. Won't the Europeans want something like N.MM.YYYY? ;-)
Grahame Grieve (Feb 01 2022 at 23:11):
only if the US want MM.N.YYYY
Josh Mandel (Feb 01 2022 at 23:16):
I don't have any strong feelings about this, but I do just want to point out that semantic versioning provides an answer to the first question you asked: bug fixes increment the patch number, and new ("compatible") features increment the minor version number. No?
Josh Mandel (Feb 01 2022 at 23:17):
If the general design is stable there is no need to go beyond a major version of one.
Josh Mandel (Feb 01 2022 at 23:17):
That said, I've got nothing against https://calver.org/
Grahame Grieve (Feb 01 2022 at 23:25):
I've found, in practice, that for the IG publisher, the difference between a bug fix and a new feature is a gray line I can't differentiate sometimes
Elliot Silver (Feb 01 2022 at 23:37):
Is there a reason to avoid three digit patch numbers?
Grahame Grieve (Feb 01 2022 at 23:50):
only aesthetic
Josh Mandel (Feb 01 2022 at 23:51):
Grey line for edge cases sure, but if 1.1.98 is the current version this suggests you never have viewed anything as a new feature.
Grahame Grieve (Feb 01 2022 at 23:52):
I certainly haven't paid any great attention to the question, that's for sure
Grahame Grieve (Feb 01 2022 at 23:54):
https://github.com/HL7/fhir-ig-publisher/releases... even new features are mostly about fixing bugs in existing features. Bugs in things that were new features to fix bugs
Last updated: Apr 12 2022 at 19:14 UTC
