On Apr 8, 2009, at 2:44 PM, Noah Gift wrote:
I do like my analogy on Python packaging though, not sure if I have shared it here or not. What would you do if, in a fictional scenario, someone kidnapped your family, and said, " In 24 hours, we are going to install your software, either on Mac, Linux, or Windows. If it works, will release your family." Would you do anything different then what you are doing now. Or are you confident that it is as reliable as possible and that it will install 99.999999999999% of the time on any system?
Wow, I think you're getting a little too worked up over this. Does it suck when an install fails? You betcha. Is any installer system flawless? Nope. Of these installers: - Ruby gems - Python egg's - RPM's - Debian packages - FreeBSD Ports/packagesEvery single one of them has failed me in the past when installing at some point. Not a single one has touched the 99% figure you speak of, even though some of the ones in that list have had *many* full-time software developers spend considerably more effort on them than the Python egg installing stuff has had.
Now, I think installing has gotten significantly better for Pylons. If you look back to after the 0.9.6 release, there was a decent amount of people reporting install issues. When 0.9.7 was released, there was a tiny fraction of the install breakage from the past. We've made a hell of a lot of progress, and none of us have full-time jobs to working on this software (unlike some of the people working on the debian/RPM installers).
Noah, you work at Weta, thats a big (and rather profitable) company that uses Python. Maybe you could get some of them to sic some full- time programmers on making the system as flawless as you speak of for the good of all python packages? :)
Cheers, Ben
smime.p7s
Description: S/MIME cryptographic signature
