That's because upstream changed the way he/she tags the releases.

I am not sure this is something I want to handle, as I am afraid it's going to open a can of worm.

Hey, so I recently learned that the python stdlib distutils module has a nice utility for comparing versions:

>>> from distutils.version import LooseVersion
>>> weird_alpha = LooseVersion('v3.0.0-alpha-2')
>>> old_normal = LooseVersion('2.5.0')
>>> weird_alpha > old_normal
True
>>> weird_alpha == old_normal
False
>>> weird_alpha < old_normal
False
>>> weird_alpha >= old_normal
True
>>> weird_alpha <= old_normal
False

oh, nice

oh, interesting

hm, this is leading to other problems:

>>> from distutils.version import LooseVersion
>>> v1 = LooseVersion('0.10')
>>> v2 = LooseVersion('v0.1.0')
>>> v1 > v2
False
>>> v2 = LooseVersion('1.1-alpha-7')
>>> v1 = LooseVersion('1.1')
>>> v1 > v2
False
>>> v1 = LooseVersion('1.0.3')
>>> v2 = LooseVersion('dev-master')
>>> v1 > v2
False

So not sure if we should change something here.

Yeah. Tricky. :(

So, solving this generally might be impossible. People could use whatever format they choose to release their software, and we're not going to be able to predict that 100%.

OTOH, the least we can do is .lstrip('v') on versions to get rid of that leading 'v' that crops up everywhere.