Quote - I think the big issue is that the python code really needs to be compiled to be faster, as code that needs to be interpreted is going to be way slower. 

The DSON importer actually uses almost no Python code. Most of the work is done by the C++ shared libs/DLLs and the geometry is loaded from OBJs as it's done traditionally in Poser. Of course we don't know exactly what's happening in the DLL and there is probably some calls to the Python Poser API to arrange things with the geometry. Keep in mind that the .duf files are compressed JSON file so they are rather verbose, plus they need to be uncompressed. Depending on the operation, this can be pretty expensive.
In general DAZ has traditionally struggled to make their code run fast, and I suspect that the DSON plugin was prepared pretty quickly with not too much concern for performance.

We can look at the timeline to get a good idea. The DSON plugin uses the new Add-On Framework that was announced by Smith Micro and myself in August at SIGGRAPH. I don't know for sure but I assume that DAZ contacted SM after that announcement and got access to the framework. 

Then DAZ released the DSON plugin in November, if I remember well. That means about three months of development. Not much, IMHO. And this is happening while they are updating Studio with other changes and bugfixes. 

That doesn't give much time to optimize things.

For what I've heard the development team at DAZ is rather slim nowadays so they are probably spread a bit thin and optimization is probably a luxury at this time.

Cheers.

Paolo

https://www.preta3d.com
FB: https://www.facebook.com/RealityPlugIn
Tw: @preta3d
G+: https://plus.google.com/106625816153304163119
The Reality Gallery: https://reality-plug-in.deviantart.com