Jeff wrote: > BTW, what about non-Python UI's? They will have to have their own installation system. I just don't know about any solutions which can effectively manage installations for tons of different programming languages, so right now all I'm worrying about is the python stuff that we have :-). If a better installation mechanism exists that can handle tons of languages and will do this generally, that would be better, but I just don't know what that would be... In general, I don't know exactly how we are going to handle non-python ui's. I guess it will depend on what language they are written it. > Okay, then you're talking about just the Python stuff (UIL and DL), because > Distutils doesn't handle C/C++. Right? Distutils only handles C/C++ that is being build explicitly for the purpose of being imported into a python program (ie. wrapped C/C++ code). There is no way it is as well suited for handling compiling big C/C++ programs like Overflow and GMS as autoconf is. > Yes, it'd be MUCH easier to re-import everything. (Getting rid of "library" > will also break a lot of things :-( ) Nah, just the import statements and xpm directories, right? I've been trying to put the directories for xpms into constants so that it would be easier to fix problems like this. > Could we use "uil" instead of "ui"? Just semantics here, but we're really > talking about a "layer". The UIL is a layer comprised of multiple UI's, and > more than one UI will go in the "uil" directory. I don't care :-). Seriously, naming doesn't mean much to me, as long as decide on something and stick with it. > Hold on a sec. My understanding is that the "site-packages" directory is > used > for Python development modules/libraries and not for applications. Have you > heard or seen otherwise? Are we considering most of Pied and the DL to be > development libraries like PyGTK? Is it really this well defined? It seems like the site-packages directory is for libraries, and the loci code is still a library, it is just that we use it and not anyone else :-). I didn't really know that their was a standard on this. By putting a piper module there we really aren't competing with anyone's namespace, so I don't really see what the big problem with it is. Fnorb, for instance, actually has it's fnidl script inside of site-packages (so that you set your $PATH to include this script inside of site-packages) so all of the code for running this script is inside site-packages. > Also, something like... > /usr[/local]/lib/piper > would be acceptable for Unix systems. Well, if we put it in site-packages, we would be putting it in /usr/local/lib/python1.5/site-packages/piper... Why is this not appropriate? > I actually had this discussion with the author of Anaconda (the RedHat > installer) when he put an Anaconda file called "gui.py" in the site-packages > directory. Well, this is kind of an ugly namespace violation (much the way orbit-python sucks up the namespace CORBA by putting its module right in site-packages), but we won't be doing this because everything will be inside of a piper module. > Don't we need a home directory for Piper anyway? I was thinking of putting > the Python code in something like /home/piper, and then a symlink can go in > /usr/bin. Why do we need a home directory explicitly for Piper? If we start doing stuff like this they we'll have to start writing our own installation scripts. <sigh> I'm just trying to make use of what is already available to do our installation for us. What kind of stuff would we be putting in this home directory that makes it necessary? Brad