Sepiida

----
Bad tag

---
fields.URI used to allow anything that is a string through. If the URI was recognized it would provide the parameters in the payload of the resource post handler. If the URI was unrecognized, like 'www.google.com' it would provide an empty dict where the parameters should be. This meant that applications had to do their own validation if they wanted to be strict.

That's dumb.

We don't have a strong use-case for allowing arbitrary strings through a URI - use the String field for that. So now we do actual validation of the URI and emit errors if we can't parse the provided value as a URI for the endpoint that the URI is set to.

Yay being automatic

---
Moved to using pytest 3 and making the requirement >= for pytest 3

---
Fix a bug with `extract_parameters()` emitting a redirect exception when passed a URL that nearly, but not quite, matches a routing rule

---
Fix a unit test that was relying on the ordering of a dict to work and had a 50% chance of failure every time

---
Update `permissions.update_filter()` to honor `permissions.always_privileged()`. Without this the `always_privileged()` context manager breaks when dealing with code, such as from chryso, that updates a supplied filter based on the permissions available to the current user. In that situation we still attempt to update the filters even though we should be privileged