Resolver and inefficiencies

[Paul Sandoz]

On May 3, 2012, at 6:44 PM, David M. Lloyd wrote:
> It's funny because version resolution is such an utterly undesirable feature.  The downsides are that it's potentially horribly inefficient, and it makes unwarranted assumptions about your runtime environment (for example, that you only want one version of something).

Presumably only one version of some artifact should be visible (and accessible) to something else that requires a compatible version of that artifact, both at compile time and runtime.

It's about the scope of one version of something e.g. per VM, per application [*]; or per dependency (for want of a better term).

Paul.

[*] Two or more applications per JVM can be isolated.

>  The upsides are... well, there are none.  Basically there's no point to having two of something installed if you can't use either one.
> 
> This feature was specified from a magical ivory tower where real use cases don't exist.
>