Implement Encode for sequences of string-like types

[tsurdevson]

This is in turn used to implement Encode for slices of string-like types. Introduce a marker trait for types that encode into ssh-strings, and implement Encode for various kinds of sequences of such types. E.g. implement Encode for Vec<String>, &[&str], etc. Downstream users could implement the marker trait for their own types and get the same behavior for slices of their own types.

Closes #314.

[tarcieri]

I'm not sure what you're trying to accomplish with this, and notably types like str and String cannot express the full range of an RFC4251 string which is a [u8]-like bytestring. This can make these types problematic anywhere you need to be able to round-trip a message encoding, such as key comments which were recently changed from a String in #360.

[tsurdevson]

I'm not sure what you're trying to accomplish with this

This was simply meant to be something I could link to as I asked a followup question to your comment on #314. I didn't have time to finish it earlier, hence the draft status and no question posted yet. In retrospect I could just as easily have linked to a commit, so I apologize for the noise of an additional PR.

I think some marker trait which denotes that encoding a message this way is actually valid for a given type would be helpful, and may even get rid of the impl conflicts

My question would have been: I'm not quite sure I understand what you mean by this. Something like this (PR)? "Encoding a message this way is actually valid" --> a trait for types that encode into an RFC4251 string?

(Judging by your answer I'm going to guess this was not quite what you had in mind.)

[tarcieri]

Oh sorry, I was missing the context. My bad.

Maybe call it something like Rfc4251String? Can you also try adding the blanket impl from #314?

[tsurdevson]

Maybe call it something like Rfc4251String? Can you also try adding the blanket impl from #314?

Happy to rename it. What blanket impl would that be? There's already these two:

impl<T: EncodesLikeString> Encode for [T] { ... }
impl<T: EncodesLikeString> Encode for Vec<T> { ... }

With some more boilerplate it can be expanded to cover all kinds of combinations of references, slices and owned types too, like Vec<&[u8]>, &[&&[u8]], Vec<&String> etc.

[tarcieri]

I was referring to this one: #314 (comment)

You don't appear to currently have a generic impl for Vec

[tsurdevson]

Only the one bounded by T: EncodesLikeString, yes. Perhaps I'm just missing the obvious, but adding impl<T: Encode> Encode for Vec<T> would still run into the impl-collision for Vec<u8>.

[tarcieri]

Yeah, that seems problematic

[tsurdevson]

I've tried to clean it up a bit and make it as general as I can without stepping on the compiler's toes. Feel free to review if you think it adds anything, or just close it if it confuses more than it helps.