I'd like to try and upstream as much of this backend as possible to the native Django pylibmc backend.

From what I can tell the main django-pylibmc features that might want to be upstreamed, are:

• Binary mode support
  • Django ticket: https://code.djangoproject.com/ticket/15815
  • PR: Fixed #20892 -- Allowed configuring memcached client using OPTIONS django/django#7160
• Username/password support
  • This requires binary mode support in addition
  • Discussion: https://groups.google.com/d/topic/django-developers/a6fACRkRCHQ/discussion
  • PR: Fixed #20892 -- Allowed configuring memcached client using OPTIONS django/django#7160
• Graceful catching/logging of the pylibmc exceptions to match the python-memcached behaviour
  • Django ticket: https://code.djangoproject.com/ticket/28342
  • PR: Fixed #28342 -- Fixed PyLibMCCache error handling django/django#8681
  • Newer PR: Fixed #28342 -- Catch exceptions when using pylibmc cache backend django/django#12893
• Support for compression
  • No Django ticket filed yet

Things that I don't think should be upstreamed:

• Automatic handling of the MEMCACHE_* environment variables
  • these are likely best suited to something like https://github.com/rdegges/django-heroku-memcacheify or https://github.com/joke2k/django-environ (the latter being possible once this is upstreamed, since it doesn't support non-native backends at present)
• Overriding the timeout handling
  • this is now redundant for the reasons mentioned in Remove workaround for setting an infinite timeout #35 / No longer override Django's native timeout parameter handling #36, so should be removed from django-pylibmc and not upstreamed

Edit: Have updated the list above to account for recent Django changes.