|
||||||
|
I've seen a few articles around the web (comp.lang.python, Doctor Dobbs etc.) about this, but a couple of hurdles I keep crashing into when using Python, which I think are holding back adoption of the language (certainly where I work and my personal use) are: 1. Distributing applications py2exe does an acceptable job on Windows, there really is nothing for UNIX (McMillan is crap, you still have dependency issues, cx_Freeze barely works) and there's no way to obfuscate your code - essentially you have to provide your source, or bytecode which is easily disassembled. I've had a similar problem at work with Perl lately, we've been wondering how we can give a machine to customers with Perl source on it, when the whole point of the software is to provide security! I really have given up trying to figure out how to distribute a Linux version of a program I'm writing, I guess I'll just bundle the source and leave it as an exercise to the user to install Python, wxWidgets, Gtk+, wxPython and the whole dependency list! 2. Speed There's something inherently wrong with the startup speed of the interpreter on Windows. I can even notice it on a 2.5GHz PC, it's OK the 2nd time you start it as Windows has the DLL cached in memory, but on Linux/Solaris it's instant. Also the CPython interpreter is constantly being made to look slow by the JITs in Jython, Psyco and IronPython. Both of these issues could be solved by a machine code compiler instead of this bloody bytecode crap, or maybe we just need a better CPython interpreter with a JIT (pycore maybe?) I'm hoping PyPy or IronPython might solve the speed problem and provide a compiler, otherwise I might switch to C# as soon as the Mono guys get Windows.Forms sorted (GTK# looks awful under Windows). |
||||||
|
|
discontent.