Pituophis

*This release contains breaking API changes- take caution when updating.*

- **Selector class has been renamed to Item**
- Globbing is fixed on Windows
- Single Items can now be sent by custom/alt handlers like strings and lists of Items can
- URL parsing has been rewritten to adhere to [RFC 4266](https://tools.ietf.org/html/rfc4266) (including `%09` for search queries)
- URLs returned by `Request().url()` do not include `:70` if the port is 70 (as this is the default port for all Gopher sites); they also do not add "/" to the beginning of selector strings (for Gopher servers that don't necessarily use this filesystem scheme)
- error.host is no longer used by default in the path/selector field of new Items (as it could be misinterpreted when looking at menu source code client-side); the field is now blank by default
- Plenty of refactoring, tidying and other good things

- TreeGopher demo added and readme updates

Default handler
- Globs no longer breaking out of public directory
- Markdown files have item type 0 (text) in directory listings instead of the default type 9 (binary)

Client
- parse_url() checks if path ends with a '/' and sets type to 1

This is actually just 0.98 but with a fixed setup.py.

Server:
- An advertised port can be set (for handler usage) for when you're rerouting the "advertised" port (like 70) to the actual server port

Default handler:
- Glob sorting (and in turn, sorting of directory listings) now works correctly, and directories are on top of files
- Checks if a file with an unknown MIME type is actually a directory before giving it that type, otherwise gives it the 9 (generic binary) type; applies to `?` pseudotype and globbing

Server:
- ***Listens on any hostname, not just 127.0.0.1.*** Fixes outside connections, which were previously broken.

- When getting the Request object of a Selector object, the query will be set if that's in the selector's field
- urllib is now used for parsing URLs
- IPv6 support (wrap IPv6 address in square brackets)