It's up to you. The program has been published as PyMCA because of the scientific use of MCA for multichannel analyzer but PyMca is more pythonic and it is what you had to type to get the program running. Due to some problems I encountered with the publisher, I have some preference for PyMca because it makes clear that it is the name of the program and it does not intend to be an acronym.
Believe it or not, Mona Lisa has more to do with the PyMca code than with The Da Vinci Code. In particular, the support of multilayered samples and of X-ray tubes was greatly influenced by the use of PyMca by the Centre de Recherche et Restauration des Musees de France (C2RMF) to analyze X-ray spectra obtained from that master piece. The results of that work were published in July 2010.
SPEC file format. That includes multicolumn ASCII, Canberra's .TK, AmpTek, and QXAS. If your format is not supported but you know how to read it, it should not be a big problem to implement it.
Starting with version 4.4.0, PyMca supports the HDF5 format. Due to its versatility, this format will progresively become the preferred input format of PyMca.
The supplied X-ray tube profile calculation tool is just for guidance and, unless you are going to measure samples that are very similar to your standards, I really doubt that you can use the generated profile without some "hand work". In any case, please consider ALL sources of attenuation between the beam and the sample and between the sample and the detector. If you aim to work at very low energies, please consider the atmosphere between the detector window and the detector itself. Some detectors are not under vacuum but under some inert gas atmosphere.
In its simplest form the procedure would consist on measuring a thin film standard and entering as matrix composition the known composition of the standard. When asking the program to calculate the concentrations using a matrix element as reference, it should give the exact concentration at least for the reference element. Then, at the concentrations tab, switch to the fundamental parameter method. Adjust the time and solid angle parameters to match those of the measurement. At that point, start to play with the flux parameter till you reproduce the same result as the one obtained with the internal reference. Once you have found a set of fundamental parameters that reproduce all your standards within the desired accuracy, you will be ready. The procedure can be/is tedious, but it is really worth the effort.
From the program point of view I would say that the bugs introduced when switching from PyQt to PyQt4 have been already fixed. In any case, if you make any batch fitting or imaging, you certainly want the latest version. Among other features you can easily find a 10 fold speed improvement ...
to start the application from the console and see any possible error output there. Of course, you will have to replace 4.3.0 by the number of your PyMca version.
cx_freeze on linux and windows. For the Mac I have used py2app. In order to make installable packages I have used the Nullsoft installer on windows and Platypus on the Mac.
Numpy is PyMca's numerical module since version 4.1.0. You will need Numeric if you try to build any previous version.