[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Good enough

Regarding my attempt to rekindle discussion and explain why
I do not believe we should standardize on any package
manager -- I think I can distill the responses into three

(1) We already decided on RPM, get over it.

(2) What you're suggesting is too hard. 

(3) We need to move forward and it will take too long to
start this discussion over again.

(4) I had the same idea but couldn't get any buy-in on it. 

My responses:

(1) Stupid answer.  If RPM is an inadequate answer, and I
believe it is, then let's revisit the topic.  If we can't
ever get to the point where we can admit to making any
mistakes, then we're not good for anything and might as
well close up shop.

(2) Lazy answer.  We have some of the most talented people
on the planet at our disposal.  Not only is it NOT too
hard, it is actually easier than trying to address all of
the problems with RPM.  All we have to do is outline the
proper procedure for ANY package installer.  Please don't
pick apart this example, because it's only meant to be a
very oversimplified outline to illustrate the point, not as
an official starting point for the new approach:

Step 1: Run pre-installation script 

Step 2: Using FHS 2.1 and standard naming conventions,
search directories for dependencies 

Step 3: Install files

Step 4: Run post-installation script

If the RPM maintainers want to add a step in there that
searchs the RPM database, so be it.  I'm not suggesting we
ban RPM.  I'm suggesting we define the rules necessary to
make package installation reliable regardless of the
package format, and regardless of whether or not anyone has
installed anything using make.  

(3) Laughable answer.  LSB is embarrassingly late already.
If you really thing the above task is too difficult for LSB
1.0, then at the very least, publicize it as part of LSB
1.1 or 2.0.  Give onlookers the confidence that we're not
just cutting corners for convenince sake, that we're fully
aware of the real installation problems and we intend to
address them the RIGHT way.  If you can't deliver it in
1.0, then lay out and publicize a credible roadmap for a
useful spec in the near future.

(4) I know how you feel.  


Nicholas Petreley   Caldera Systems - LinuxWorld/InfoWorld
nicholas@petreley.com - http://www.petreley.com - Eph 6:12

Reply to: