Closed jas88 closed 2 months ago
Unfortunate that they have two variants. It's so much easier to just sign the assembly even for modern .NET use and let it ignore the fact. If I remember correctly the advice (at some point at least) was that sign the assembly and have the public key public in repo (don't expect it to be a secret).
@lahma yes, signing in itself is generally useless but harmless, and they've added a "PublicSigning" option to simplify that too - the problem is that Mathnet.Numerics has two different packages with conflicting names paces, and they specifically advise use of the unsigned variant except for the handful of cases that actually have a use for signing.
.Net Framework cares about legacy "Strong Naming" which is why #785 was opened in 2022, but it's not useful for Core - this PR makes it conditional, so Framework and Standard 2.0 users get that signed variant, Standard 2.1 and Core 6 will get the normal package instead resolving #1365 without upsetting legacy people.