[Percent encoding] [Question] Do the percent encoding functions really need `'static` `AsciiSet`s?

[Scripter17]

Currently, percent_encode and utf8_percent_encode require the AsciiSet to be 'static, despite already having a lifetime bound 'a on the input string.

This is very rarely an issue, but as part of a URL cleaning tool I'm making I allow users to choose various common AsciiSets from the URL spec and javascript. The API I've designed is an enum of names corresponding to those AsciiSets. For flexibility I'm considering adding a Custom(#[serde(some_glue_stuff)] AsciiSet) to allow users with niche cases to not have to bodge stuff. This would require the getter function to go from get(&self) -> &'static AsciiSet to get<'a>(&'a self) -> &'a AsciiSet, which would make it incompatible with the percent_encode and utf8_percent_encode functions.

Fortunately the cases where this is needed are rare and perfectly fine for me to just upstream with a new name (which also avoids bloating the enum from 1 byte to 17 bytes), so it's not a big deal if the functions can't be changed to accept &'a AsciiSets. It's just not clear to me if or why they can't.

[Manishearth]

My suspicion is that they do not, would entertain a PR that changed the API; though I'd like @valenting to confirm.

I was also wondering about this when the API was added but I was particularly busy then.

[valenting]

I think changing the signature to make AsciiSet bout on 'a would be a breaking change, but adding new methods that accomplish this might be OK.

[Manishearth]

Why would it be breaking? It's not s trait.