I'm +1 on having this committed.
I'm using this on a daily basis. I have 12 types of services sending 27 different Thrift structs back and forth with no problem. I have been tracking this patch and Esteve's updates to it over the last 4+ months. I've also spent time digging around in the generated Python, and it looks fine to me.
This code is already in use by several others who are also tracking it as Thrift changes and Esteve releases new patches to track changes to Thrift trunk.
Almost all of the changes in this patch are isolated from the rest of Thrift. This is because the patch adds an option to the compiler to generate Twisted-specific Python. The Python (ttypes.py) that is emitted for classes implementing Thrift structs is identical to the Python ordinarily produced. The compiler changes are isolated in an if clause that tests whether Twisted output has been requested.
It would be great to get this committed. It only affects the people using Twisted, and while there may be future modifications to this functionality, it would be better and considerably less work for all concerned if this patch were in trunk. We can and will enhance the functionality, but it's much easier to do so in the context of a trunk which already has the underlying functionality provided by this patch.
Because the basic functionality is already known to work (see above), patching one's copy to include future minor changes is likely optional to people already trying to work with Thrift and Twisted. Instead we can get on with other things and pick up future enhancements in a later Thrift release.
Disclaimer: I work with Esteve - though he does the Thrift stuff entirely independent of things I work on. I just use the product of this patch.