[gobolinux-users] why gobo do not want to adopt any existing package manager?

Mikolaj Habdank mecat at wp.pl
Fri Jan 21 19:31:44 GMT 2005


I was wondering why gobo do not want to adopt any existing package
manager? Recipes are easy and fun, but has no power needed by developers
and user. I would like to see this features:

profiles - an easy way to switch between diffrent compilers (3.3, 3.4
etc) - it is very important for working with amd64 because there are
diffrent problems with diffren compilers and i cannot say that any one
of them is better.

md5 or another checksum for scanning installed packages (are they
correct?), not only security scans but also scans for broken files (for
example while instaling other packages - look at libGL)

deps deps deps - gobo need perfect system for scanning dependencies with
multiple alternatives - for example package need sound backend but we
have alsa, oss, jack and others (see poldek in pld as a good example)

sandbox - i have to be sure that all packages which are deps will be
build correclty before i will install new package. even if i have
recipes for few packages i cannot be sure that they will compile on my
gcc and my glibc and my... etc.

rebuilding installed programs database - yes it is possible in gobo
becouse we have each package in diffrent dir! this can be better than
portage because if portage is corrupted the only one way is to rebuild a

gobo need binary packages - we should base on pld repo - each
arhitecture have special dir with packages - 686 is not a good way. it
is not best optimisation and legends about 5% speed increment is stupid.
Try codecs and stop talking bulshit. Otimization is very important if we
want from machine as much as we pay for.

I know that "my gobo" is far diffren from GoboLinux (but it is not my
fault that i can not build it normaly on amd64) but i know that recipes
are too simple.


More information about the gobolinux-users mailing list