Update semver.md

[ajbogh]

Clarified points 4 and 5.

Fixes #490

[jwdonahue]

If my product is not "ready for the public to use", why would I burn through SemVer version numbers at all? If I am not publishing the product, a build number should suffice.

[ajbogh]

@jwdonahue I appreciate your comment but I'm not entirely sure how it relates to the PR. Regarding version numbers, you're welcome to use any format but this change only entails the details of items 4 and 5. Other items describe using notations such as "-prerelease" and build numbers. In fact I'm using a package that uses "canary-xyz123" as a version number.

Again, this PR is only to clarify and improve items 4 and 5. There is still some leeway given with the verbiage, such as using the words "It is not recommended for the public". Just because someone does not recommend the usage doesn't mean it's off limits. We just want to point out that certain guarantees in code quality are not provided.

[jwdonahue]

@ajbogh wrote:

We just want to point out that certain guarantees in code quality are not provided.

That's already obvious in the current spec. It does not require added emphasis.

[alexandrtovmach]

Closing & re-opening to trigger CI

[adamralph]

Something feels off with this. I'm not sure that recommendations for and against public use should form part of SemVer. After all, if something is not recommended for public use, why publish it?

[alexeyzimarev]

There are quite a few popular projects that avoid going 1.0 for the sake of flexibility to make breaking changes. Kubernetes was on 0.X for a while and used in production by hundreds of thousands. Hugo is another example, it is on 0.81 now and millions use it.

[adamralph]

Something feels off with that.