| 27 Mar 2020 |
| 11:47:40 |  Cadair | Our deprecation policy being we must raise a dep warning in at least one LTS before a breaking change is made (for a given definition of breaking and where possible) |
| 11:47:56 | Cadair | Sometimes you just have to break stuff |
| 11:49:09 |  Jacob Tomlinson (Slack) | I feel like from the way you've described it there is a subtle difference between this and semver. But I'm just not quite seeing it. |
| 11:49:15 | Jacob Tomlinson (Slack) | Why is that different to semver? |
| 11:49:39 | Cadair | The version number isn't used to indicate API changes |
| 11:49:46 | Cadair | our deprecation policy is |
| 11:50:14 | Cadair | we can increment the major version number and actually not break anything |
| 11:53:41 | Jacob Tomlinson (Slack) | What is the motivation behind that? |
| 11:54:24 | Jacob Tomlinson (Slack) | Incrementing the major version without breaking anything I mean. To draw a new line in the sand in terms of support? |
| 11:56:13 | Cadair | yeah |
| 11:57:47 | Jacob Tomlinson (Slack) | How do users know whether a new release is safe or not? Trial and error? |
| 16:52:20 |  iguy | Why not use semver standard that everyone in world knows and understands? |
| 16:54:13 | Jacob Tomlinson (Slack) | Well that was the question which started the discussion. There are valid reasons why semver isn't perfect. Personally I think people should be disciplined and use semver, but that isn't always practical. |
| 16:55:26 | Jacob Tomlinson (Slack) | The discussion link that Cadair shared is an interesting read. I don't think I agree with it. But it's interesting to hear counterpoints. |
| 16:56:54 | Cadair | depending on what your package is I strongly disagree with "everyone knows and understands" |
| 16:56:55 | Cadair | most research scientists, do not and will not understand semver for example |
| 16:57:06 | Cadair | * most research scientists, do not and will not understand semver for example |
| 16:58:54 | Jacob Tomlinson (Slack) | I agree with Cadair here. Many folks who write software are not software engineers. We can't expect them to be as disciplined with it. |
| 16:59:32 | Jacob Tomlinson (Slack) | Lots of code, like research code, doesn't need the same rigor as production code that a company would be built on. |
| 16:59:46 | Cadair | I am more thinking use software like sunpy / astropy / mpl etc |
| 17:00:09 | Cadair | the closer you get to the end user (into more specalist libs) the less value semver adds |
| 17:00:54 |  SolarDrew (Old) | In reply to @_slack_opsdroid_U5MK1BW83:matrix.org
Lots of code, like research code, doesn't need the same rigor as production code that a company would be built on. would be nice for research code to find a middle ground between "the same rigor as production code that a company would be built on" and no rigor whatsoever... |
| 17:00:20 | Cadair | * the closer you get to the end user (into more specalist libs) the less value semver adds |
| 17:01:34 | Jacob Tomlinson (Slack) | In reply to@SolarDrew:matrix.org
would be nice for research code to find a middle ground between "the same rigor as production code that a company would be built on" and no rigor whatsoever... Very much so! |
| 17:02:00 | SolarDrew (Old) | In reply to @SolarDrew:matrix.org
would be nice for research code to find a middle ground between "the same rigor as production code that a company would be built on" and no rigor whatsoever... present company excepted 😛 |
| 17:02:50 | Jacob Tomlinson (Slack) | Personally I think if you provide software for others to depend on you take on the responsibility of providing stable and we'll engineered code. Providing a public API is a step change in how much rigor you need to put in. |
| 17:04:36 | Jacob Tomlinson (Slack) | But again I think my thoughts here are a little biased. |
| 17:06:50 | Jacob Tomlinson (Slack) | In reply to@SolarDrew:matrix.org
present company excepted 😛 I think task and environment make a huge difference. If I'm building a package I behave very differently to when I'm doing some data analysis in a Jupyter notebook. |
| 17:07:43 | Jacob Tomlinson (Slack) | In reply to@_slack_opsdroid_U5MK1BW83:matrix.org
I think task and environment make a huge difference. If I'm building a package I behave very differently to when I'm doing some data analysis in a Jupyter notebook. I'm a zealot for strong software engineering practices. But I'm also terrible at version controlling my notebooks. |
| 22:36:44 | |  1o11 joined the room. |
