I would say the case is for us to define, no?

To me, upper case looks better to start a sentence/paragraph. That's the case with the rest of the text and what I am most used to seeing on Git/GH.

That would be ok too, if that's what we decide. (Git describes these as being like RFC822 headers, so from Git's point of view, they should be case-insensitive.)

I am not opposed to case insensitive, but I would like to understand the motivation. If we allow that, we should update the corresponding contribution rules.

I noticed this when looking over the change again and saw that our code was checking for "Sentence-case", but the tests used "Title-Case", and so they should have been failing. I don't have an opinion about which style(s) we accept though. I just modified the code to match the now-fixed test.

What was the sneaky error exactly?

The test cases mixed string concatenation and commas incorrectly, so we weren't testing what we expected. For example:

            "[msg] Subject\n\n"
            "> This body line (verbatim text) is clearly over 72 characters long[...]",
            "",
            "> This body line (verbatim text) is clearly over 72 characters long[...]",
            "",
            "Regular text"

From reading the surrounding context, that should be a single commit message to test, but it's actually five:

1. ""[msg] Subject\n\n> This body line (verbatim text) is clearly over 72 characters long[...]"
2. ""
3. "> This body line (verbatim text) is clearly over 72 characters long[...]"
4. ""
5. "Regular text"

Got it, thank for clarifying @jimporter! I have a slight preference for Sentence case, but I don't mind changing it if you like, provided it's all in sync with CONTRIBUTING.md.

Got it, thank for clarifying @jimporter! I have a slight preference for Sentence case, but I don't mind changing it if you like, provided it's all in sync with CONTRIBUTING.md.

We could also require Sentence case, but I think if that's our decision, we should make a separate rule for it so that we can report a more-specific error message. (Currently, we'd only report an error if the commit message contained a header like this that broke some other rule, like our line-length rule.)

You're right, for some reason I was under the impression that the rules implied case today.