handling encodings

Here is my (very early in the morning) assessment about what we should do regarding scanning the file and encodings (related issue: https://rt.cpan.org/Ticket/Display.html?id=78434)

• we need different interfaces for new-from-string, new-from-file, new-from-handle. new-from-string will assume decoded characters. new-from-file takes an encoding parameter, and uses binmode(":encoding($encoding)") on the handle. new-from-handle assumes that the caller has already set encoding layers appropriately.
• new-from-file may not have been provided an encoding parameter. in this case we start reading assuming ascii. If we encounter a 'use utf8;' directive, we back up and start over with the correct encoding applied. We will also flag a warning (error?) if we encounter a 'use utf8' directive when we were explicitly passed a different encoding parameter than 'utf8' (UTF-8?).
• pod can have a different encoding than the code. when starting to read pod, and we don't already have an encoding supplied (can we assume that new-from-string is decoded and therefore exempt ourself from this step?) and we encounter an '=encoding ...' directive, we decode the extracted pod lines as they are read. Test cases need to consider corpus examples where the code is in a different encoding than the (interleaved) pod (can this happen in any situation where the code is not ascii, which is utf8-compatible?)

Am I missing anything?