Reverted to published language RE build metadata #182
Conversation
"Build metadata SHOULD be ignored" instead of "Build metadata MUST be ignored".
|
Should we increment the number to 2.0.1 in both places? Or is this just a clarification of the 2.0.0 intent? Anyone have thoughts? |
Added a sentence to specify that an all-digits pre-release identifier with a leading zero is not a valid alphabetic identifier. Nor is it, according to the existing documentation, a valid numeric identifier.
|
Changing from "MUST" back to "SHOULD" (e62c5f2) is actually a net change of nothing, when compared to the published version on semver.org, so I wouldn't consider that a patch change. |
|
Ack, I'm confused. Why are we changing this? I thought we were going to change semver.org to match semver.md here. |
|
I think |
|
I'm sorry if I've confused you! The markdown document claims to be SemVer 2.0.0 but it is inconsistent with the published document at SemVer.org. Therefore, given that it is a deviation from the published interface, it MUST be version 3. That's why I reverted back to "SHOULD" in order to maintain compatability with the existing, published interface. Since this is a change back, no increment is required. The clarification about treating "0123" being an invalid alphanumeric identifier was ambiguous in the published document but it matches up with the (unpublished) BNF you pulled a while ago. Therefore, I think it can safely be considered a bug fix, hence 2.0.1 As for submitting a change on SemVer.org rather than here, the document itself points to https://github.com/mojombo/semver/issues. Is this not the right place, then? |
|
Ah, the document here is the work in progress one. The one on Semver.org is 2.0.0 and corresponds to this commit: https://github.com/mojombo/semver/blob/7c834b3f3a4940d77ab593bc32583004d6a426a9/semver.md We probably need to update this one to as the in-progress 2.0.1 immediately to avoid confusion. |
|
Did that make sense? So no PR to semver.org is necessary yet. We're still working on an update to 2.0.0 that we'll tentatively call 2.0.1 unless we introduce a breaking change or a "new feature". |
Not your fault. I'm doing 10 things at once and I thought I was on semver.org when I read this PR. Sorry for the confusion on my part! |
|
Sorry to keep on, but I'm not convinced you're right. If you don't change the wording of section 10 back to "SHOULD" then you already have a breaking change in the next version of the spec which will require the version to become 3.0.0 rather than 2.0.1. Also, the other thing is a genuine bug because it's an ambiguity in the spec. I'll leave it with you though. |
|
Sometimes following SemVer is a pain. Even when you are semver. Ok, perhaps it should be 3.0.0 |
|
Some implementations are following the SHOULD and actually provide sorting possibilities based on meta data (I guess if they follow a given pattern) to cover certain uses cases. So it must be changed back to SHOULD for now, and might be worth asking for use cases among misc package managers and similar software for inclusion in version 3.0. |
|
Above it seems to be assumed that SemVer itself should follow SemVer restricting the ways of resolving this ticket. If SemVer had followed SemVer the existence of more than one version of 2.0.0 would mean that it would seem it is forever non-compliant and reverting the changes wouldn't help, see - #229 As for the specifics of meta-data in precedence http://semver.org/ item 11 states precedence MUST only be determined by specific data excluding meta-data, and item 10 states that meta-data SHOULD be ignored. Thus any tool that uses meta-data in comparison is not correct according to item 11 - regardless of the contents of item 10. |
alexandrtovmach
added
the
spelling/grammar
|
Closing & re-opening to trigger CI |
alexandrtovmach
added
extend
"Build metadata SHOULD be ignored" instead of "Build metadata MUST be
ignored".