implemented features (Re: Decision time....)

Bill Hacker wbh at conducive.org
Tue Jun 5 16:33:06 PDT 2007


talon at lpthe.jussieu.fr wrote:
km b wrote:

On 6/5/07, Michel Talon <talon at lpthe.jussieu.fr> wrote:

*snip* (64 bit issues)

whatever he sees fit for his project. My own prejudice is that having a
good installer and a good package management system (including a good
upgrade mechanism) is *extremely* important, far more than any kernel
nicety. This is clearly the Achilles heel of all BSD systems. Beleive it or
not, the talent to develop such tools is completely orthogonal to the
talent to do kernel programming.
I don't see a lot wrong with pkgsrc.  Nor any 'easy' (read 'with available 
resources') way to improve upon it.

Though I personally still prefer the scope of the FreeBSD ports tree, we all 
know the huge 'count' of ports has a lot of redundancy and no small amount of 
plain useless or broken junk. So pkgsrc's smaller count is not really all that 
much less w/r what can actually be *utilized*.

As far as 'good installer', I experiment with anywhere from half a dozen OS in 
any given year.

My all-time favorite intaller - 1970's to this very week - remains the legacy 
FreeBSD 'sysinstall'.

Sorry 'bout that, I know a lot of work has been put into DFLY's own installer, 
and more coming - but given the choice I'd prefer 'sysinstall' even over the 
best-of-breed Loonix variants.

It is always readable, local console or remote ssh, gives me the flexibility I 
need - not only for FreeBSD, but to set up HDD for DragonFly and other OS'en as 
well. It doesn't use fonts that eye-confuse, either, and I've gotten 'adjusted' 
to the way one exits/navigates back up the tree, though that *could* be improved.

Whatever the direction is to be, having a solid kernel and core is far more 
important that a slick installer to put in place a 'still needs work' OS.





More information about the Kernel mailing list