A large part of why Python 3.6 is a better port target than 3.0 is because they slowly and silently stepped away from their very strong opinions about "what is text", and "what is bytes", and what each of those things meant. Even back when the model was debated, it was clear that the operations the core maintainers cared about (manipulation of individual codepoints [0]) was a dead-end, as the Unicode Standards committee was just about realizing that codepoints weren't an ideal unit, and grapheme clusters were added to the spec. (my dates might be a bit off here, it's all a blur I've mostly forgotten)
Python 3.0 was released as basically a completely nonfunctional piece of software. You basically couldn't write anything that touched a binary file in it. The built-in zipfile module, I remember, crashed if you tried to store or load a binary file stored inside a .zip, the only tests at the time stored .txt files, and the bug wasn't fixed for a long time. I remember others having troubles with the email and mime modules, though I didn't have any code that worked with those, personally.
I had left the community by that point, but I have heard tales of the mountains Armin Ronacher had to move in order to add the u'' prefix back to Python 3; the core team was against it, instead believing that if 2to3 wasn't working for people, it should be fixed. Alice Bevan–McGregor spent years making a version of the WSGI specification that worked with Python 3's text model. We honestly could have shaved 5-6 years off of that disaster if they had just listened to the community; pretty much every opinion we had shared eventually came true, and the Python 3 text model these days is in a very different, and much healthier, place.
[0] One of their major strongholds was that they really needed O(1) indexing of codepoints (it wasn't clear why this was a priority, and I remember arguing that wanting this is a sign of poorly written code). I think in Python 3.8 they finally caved on this, with a PEP and an implementation that can use UTF-16 or UTF-8 internally.