[Distutils] thoughts on distutils 1 & 2 (addendum)

has hengist.podd at virgin.net
Wed May 19 12:14:57 EDT 2004


Couple more thoughts to add to yesterday's:

- Decoupling DU build procedures from DU installation procedures also 
implies delegating control of building and installing to separate 
scripts (e.g. 'build.py', 'install.py'), rather than having a single 
'setup.py' script doing double duty. Superficially this may sound 
like it's creating more(!) work for module developers, but bear in 
mind my goal of making these scripts more generic so that in most 
cases the module developer can simply [re]use existing ones rather 
than have to code new versions each time.

- Regarding the "human-readable flag" aspect of setup.py, as this 
does cause some concern... In an _ideal_ world, the absence of a 
setup.py script would simply indicate that a module could be 
installed via a generic installation process. This _not_ being an 
ideal world, however (i.e. non-DU-compatible modules also lack a 
setup.py script, making it hard to tell the two forms apart), there's 
no reason a standard 'install.py' script couldn't always be included. 
Also, removing metadata from setup scripts means that most of the 
time 'install.py' will be a completely generic script that can be 
automatically added to the .zip at build-time; one less thing for the 
module developer to have to do themselves.



More information about the Distutils-SIG mailing list