fwiw, I have a Python 3 branch which still has a lot of failures during py.test, but isnt too far away from being usable, if only to prevent further regressions.

hmm. seem travis isnt providing builds for PRs.
Here is my travis build of this commit: https://travis-ci.org/jayvdb/spreads/builds/120844859

I vaguely remember there being a reason for those explicit bytestrings, but I honestly cannot remember any more :/ But the tests are all passing, so we should be good :-)
Weird about Travis, I'll have to check the configuration, thank you for pointing that out.

However, which Python 3 version are you targeting with your branch? Starting with Python 3.4 the bytestring literal syntax is supported again.

Thank you for your work, Python 3 compatibility has been on my radar for a while, glad to see someone taking the initiative on it :-) If you notice some of my other packages that spreads depends on (chdkptp.py, gphoto2-cffi, jpegtran-cffi, etc) having trouble with Python3, please open an issue on the relevant project and I'll take care of it.

Im targetting 3.4+ initially.

Good to know you're keen, and willing to merge preliminary patches.

Will push a working .travis.yml with 3.4 in the matrix asap, so it is on everyones radar.

If it's not too much hassle for you, I'd prefer it if you collected all Py3-related changes in a single pull request that I can review from time to time and then merge once all the tests are passing on Python 2.7 and Python 3.4.

I'd prefer to get a few of the basic underlying build problems fixed with early PRs. e.g. this one and #224 , where you'll know better than I can if the basic fix is probably ok, but real world testing is also going to be needed.

As for the actual code changes, I'll do a clean series, and you can identify any that you'd like me to submit as PRs early, if any.