DragonFly kernel List (threaded) for 2005-04
Re: Stable tag will be slipped Sunday and release engineering will begin Monday
On 04 Apr 2005 18:58:05 GMT
Rahul Siddharthan <rsidd@xxxxxxxxx> wrote:
> Bill Hacker wrote:
> >Chris Pressey wrote:
> >> On 04 Apr 2005 07:11:25 GMT
> >> Rahul Siddharthan <rsidd@xxxxxxxxx> wrote:
> >>>Similarly, if one doesn't upgrade the system for 10 months,
> >"apt-get >>upgrade" or "apt-get dist-upgrade" just *works*, you
> >rapidly have an >>uptodate system.
> >> One thing I'd like to find out is whether this is due to
> >technological > or QA factors. I suspect it's mostly the latter,
> >although there are > probably some technical aspects that foster it,
> >too. >
> >> -Chris
> >It is largely QA and a firm hand *about* QA in Debian's case.
> It's both QA and the technology.
I believe that... I guess what I was thinking was, the only part we can
"borrow", is the technology. The QA comes down to sheer amount of
The question is, _what_ is that technology? I've only used Debian a
tiny bit, so I can only make educated guesses at this point.
> >The traditional *BSD methods, be they port, package, or direct build,
> >are far more forgiving and resilient - and generally easier to
> >troubleshoot, fix, or work around to keep current.
> Bill, I don't think you've actually used Debian. Try it and see.
> As for me, "resilient" is not the word I would think of when
> considering the FreeBSD ports tree. I remember at least 2 serious
> snafus (involving libpng and gettext), and have forgotten dozens of
> minor hiccups, in my 4-5 years of using FreeBSD. Usually nowadays,
> when such a major disruptive upgrade happens, detailed portupgrade
> instructions are posted on the lists, and things still seem to go
> wrong for users.
My main gripe with ports is that it is so open-ended (port creators can
basically do whatever they like so long as it's possible to describe it
in a Makefile) and badly insulated (most of my problems have come from
something bad unintentionally set in my environment and/or make.conf,
for example.) This makes it difficult to do QA on.
My impression (and it's only an impression) is that the apt stuff is
more predictable/better regimented, making it easier to do QA on.
Perhaps it goes without saying, but I'd much prefer a system for which
there is correspondingly less *need* to "troubleshoot, fix, or work