I'm not sure who our resident PS expert is.@jkseppan?@mdboom - any ideas?

This looks fine to me.

Is there any documentation needed for this change? Is theps.usedistiller== False new functionality?

I don't think it's new functionality, only that it will now write directly to a file, rather than to a temporary and then moved. I suppose if someone was relying on the old behavior that may be a problem -- but this change I think ultimately makes things less surprising.

To clarify, you cannot write "ps.usedistiller = False" in the matplotlibrc. When I said "is False", I meant if the Python variable is False, which it is ifps.usedistiller is omitted from the matplotlibrc. I have changed the Git commit message to clarify.

I don't see how someone could be broken by this change. There is no noticeable difference between the old and new behaviour unless you are watching the temporary directory for files being written and quickly deleted. As far as I can tell, the only difference is that it works even if there is no writable temporary directory.

I was thinking more about a difference caused by using a stream vs. a random access file handle. (We've had similar issues with the PDF backend). But it seems that the postscript backend doesn't exhibit such problems. I think this is good to go.