Comment 10 for bug 201630

It is actually a very good question. Your are right. I was just looking at the WMF import 'native' code and it seems like this is a proper implementation based on the Microsoft specs added in Nov 06 maintained by Ulf Erikson with a last update in Nov 07. However it is difficult to see which implementation is better (internal or uniconv) in terms of performance and compliance.

I am trying to look into extensions that require external dependencies under Windows. It seems like we might need a bit different approach as you are right the mentioned change would break other systems. So far the approach to windows extensions seems to have been to include all dependencies with the distribution (like perl and python) and if the same was followed with uniconv this would probably be the easiest approach.