Support for other versioning schemes

[NetOpWibby]

Hi y'all.

I favor ChronVer over SemVer and it's irritating to have my flavor of versioning mangled into SemVer. At the very least, I would appreciate a flag that'd let me bypass SemVer checks for certain packages (i.e., my own).

This would be a user-configurable thing, with SemVer as default to keep 99% of the userbase satisfied.

[ljharb]

That seems like a ton of complexity to add. How many packages use this versioning scheme? How popular are these packages?

The entire npm ecosystem functions on semver; it seems like if you want a different versioning scheme, a different ecosystem would be more appropriate.

[NetOpWibby]

So npm cannot ever support another versioning system? Ever?

[ljharb]

I don’t see any way to do so that wouldn’t cause compatibility issues with every extant package - but more importantly, i don’t see any value to be had by doing so.

[NetOpWibby]

I am not suggesting replacing SemVer. At all. Why would compatibility issues occur?

[dominykas]

@ljharb I don't think I agree with your position the people should go play elsewhere.

I also don't think it's true, by far, that semver is the only versioning schema that's used by the ecosystem published in npm. TypeScript explicitly does not follow semver.

I think there is value to be had in allowing packages to specify explicitly that they are not following semver. They should, but they don't have to.

I also think there is value to be had in broadening and accepting more people and packages into the ecosystem. Of course, in this specific case, I'm sure there's implications, because there's lots of existing assumptions about what are the allowed version specifiers. I'm also sure that it's possible to solve this. And I'm most certain that it's unfair and unjust to tell people to go look for another ecosystem, especially because they just might do that, and that would be a loss.

At the very least, the ideas are worth discussing.

[ljharb]

There’s two aspects here: following semver, which certainly not everyone does, and using semver-compatible versions, which is the part that i don’t think can be changed.

[dominykas]

let me bypass SemVer checks for certain packages

Which checks do you mean by this specifically?

[NetOpWibby]

What prompted me opening this discussion was an issue that arose with npm telling me a version number was invalid. I'm not quite sure how I managed to publish a weird version to my Verdaccio instance a few months ago but a fresh npm i and explicitly choosing a version to install got me around it.

[dominykas]

What was the invalid version?

[NetOpWibby]

2021.2.1-1.1. The one before it was 2021.2.11, which is what SemVer converts it to from 2021.02.11.

[dominykas]

Ah, right, the leading zero.

I personally don't see a major problem if leading zeroes were allowed by npm in some form (they are explicitly forbidden by the semver standard, though).