[gpsd-dev] python2 and absolute paths

[Eric S. Raymond]

I've had time to think about Greg Troxel's position that the right
things to do is to have GPSD configure an absolute path into itself at
build time.  I have concluded that it is ridiculous - not because of
the complexity it would add to the GPSD build alone, but because of
what requiring a parallel maneuver would do to the deployment
complexity of *every single distributed Python program everywhere*.

Linux distributions seem to have settled on a convention of having a vare
'python' of uncertain version, a 'python2' that goes to Python 2 and a
'python3' tha goes to Python 3.  For the price of two symlinks in system space,
this means a Python program can be reasonably sure its #! like will invoke
something reasonable for.

The alternative would require either that every single Python script be passed 
to
an explicitly-chosen interpretere by hand (e.g. "python2 foo.pty") or have
surgery done on its hashbang line.  That's an absurdly large amount of
effort compared to two symlinks.

Accordingly, my conclusion is that BSD (and whoever else isn't
providing python, python2 and python3 links in the standard $PATH)
should be told firmly to fix that bug.  Greg Troxel, take note.
-- 
                <a href="http://www.catb.org/~esr/";>Eric S. Raymond</a>