Yes. I'm bringing this up because it breaks certain OAuth 1 clients against Bitbucket.
In OAuth 1, requests are signed, including the request URL, but the RFC-5849 does not mandate any pre-processing of the URL. For several OAuth clients (including requests-oauthlib and python-oauth2) that means they compute the signature over a string that contains "%3A" instead of ":".
On the server however, the request path automatically gets unquoted before it hits the middlewares and views. As our OAuth layer is a middleware that reconstructs the signature, it ends up computing over ":", yielding a different signature than the client, breaking authentication.
This might be addressable by changing these OAuth clients to perform unquoting on the path segment, but a better solution would seem to make urlresolvers.py:RegexURLResolver respect the reserved characters for path segments and not escape what does not need to be escaped.
I'll follow up with a pull request, unless there are string feelings, or unwanted consequences of that approach.
Cheers,
Erik