Should patterns be automatically encoded and normalized?

[wanderview]

The URL() constructor does a number of automatic transformations to make a URL canonical. For example:

1. Encoding non-ASCII characters using percent encoding. Hostnames may be puny encoded.
2. Flattening out /./ and /../ in pathnames. Developers often rely upon this when they begin relative pathnames with ./.
3. If a port number is provided and it matches the protocol's default port, then its changed to the empty string. For example, new URL('http://foo:80').port evaluates to the empty string.
4. Probably more...

Currently URLPattern only performs these normalization steps on the input to test() and exec(). Patterns passed to the URLPattern() constructor are not normalized in any way. This means that developers must write their patterns to target canonical URL output manually.

We could possibly perform some normalization for patterns, but not in all cases. For example, if a pattern contains ([é]) we cannot safely expand this to ([%C3%A9]) since it completely changes the meaning of the regular expression. Any normalization would like only take effect outside of regular expression groupings in patterns.

What do you think? Would some normalization be preferable even if its not able to be applied universally?

(Note, this is also discussed in #33, but we would like to collect community feedback in the discussion and then summarize in the issue.)