Dwayne Bent

> If mod X depends mod A which is modified by mod B, then X can't depend on mod A, and we can't actually provide mod A since we can't...

Ah, now I understand. I should have mentioned that in the OP. While it's true that a modification could drastically alter the behavior so that mods that `depends` `ModA` would...

And yes, in that case `ModX` would not be installable because it would `depends` `ModA` which `conflicts` with `ModB`. That seems like reasonable behavior to me.

I think my proposal is actually much simpler than the impression I gave, hopefully the explanations of the puzzles will clarify how I intend this to work. At a high...

That's part of the bit about a mod with `modifies` really having multiple versions. We'd have to keep track of the version of `NotSoAwsomeMod` we have installed _and_ the version...

- Very strongly in favor of interfaces in separate files. It's a psychological thing, but keeping the interface in the same file as the main concrete implementation leads to rather...

Well there are a few ways to do DI, but the most popular I've seen has always been constructor injection for the reasons you state. An object declares its dependencies...

@pjf You pretty much got it. Regarding .NET 3.5 vs. 4.5: - There are other options for IoC containers, that probably work with .NET 3.5, including older versions of Autofac....

Reopening with backwards-compatible twist: Add a boolean flag to indicate whether or not the `version` property is semver compliant or not: ``` json { "spec_version": "1.x", "identifier": "SemanticallyAwesomeMod", "version": "0.1.3-beta2+abcdef",...

> The problem with this solution is of course that some of us have a strong tendency to add a lot of metadata by e.g. mending broken submissions from the...