S3 does not require normalization of S3 object keys and uses strict byte
comparison of object keys, not equivalent unicode character comparisons,
to store and retrieve objects. This means that storing and retrieving objects
with fog would cause the objects to be inaccessible by other libraries,
languages, and systems that don't normalize object keys. Given that there is
no benefit to normalization, except perhaps reducing byte count of object
keys, it ought to be removed.
Since `fog-json` was extracted from `fog` it means if you rely on:
require "fog/provider/compute"
`Fog::JSON` was not defined although it was supposed to be a supported
use case.
This adds the requires for JSON or XML based APIs based on a quick scan
of each provider to declare the required part.
AWS seems to be relying on both. Which is nice.
Providers relying on native code dependencies
