[Mageia-dev] kernel 3.0 is a big mistake in cauldron

Sat Jul 16 21:28:00 CEST 2011

> Seriously, look in a mirror and judge yourself, not others.
+
+OK, updating from 2.6.38 to 2.6.40-0.rc7 aka 3.0.0-0.rc7 was the right
+thing.
+
+Satisfied now?
+
+There seem to be several standards when judging what to do in an unstable
+distro.
+
+In a released distro, the common rule is _not_ to update anything, unless
+the minor updates stop and there isn't any other way to provide a security
+fix. A next major (or, sometimes, even a next minor) update only comes with
+the next distro release. Backports are possible, but optional. People
+needing newer versions of specific packages either are building them
+themselves from upstream (quite rare), or they're using 3rd-party repos
+(e.g. PPAs).
+
+So, sorry to repeat: in a _stable_ distro, even upgrading to the next
+_stable_ release is usually _forbidden_. (Fedora is kind of an exception
+here, but passons...)
+
+In something like Cauldron them, there are several degrees of possible
+freedoms:
+
+i. Upgrading packages to the next _stable_ version, like the recent
+LibreOffice update.
+This is the first degree of differentiation from a stable/release distro.
+
+ii. Upgrading packages to the next _beta/RC_ version, like it's usually the
+case with KDE, GNOME (and rightfully so).
+This is the second degree of differentiation from a stable/release distro.
+IMHO, this should be used with care.
+When you say "caludron is by definition unstable, we don't need an extra
+testing stage", you implicitely state that Fedora Rawhide's policies are
+stupid and only yours is a valid one.
+
+iii. Upgrading the _kernel_ (or the system and session manager, or other
+_critical_ system parts) to the next _beta/RC_ version, which IMNSHO is not
+the right thing to do, unless there are 2 different metapackages
+(latest-stable-kernel and latest-kernel), so that people who are using a
+cooker/cauldron/rawhide/unstable distro be able to choose the degree of risk
+they're willing to adopt:
+a). risking everything;
+b). risking mostly application/DE breakage, yet having a reasonable degree
+of confidence that the system as a whole is not really broken except in
+extremely rare cases.
+
+
+Of course, your distro, your policies.
+
+
+OTOH, I've had in the past the proof that FLOSS developers typically lack
+common sense. A few examples:
+
+1. Arch Linux developers can't understand that developing a few scripts (a
+la Slackware, not a la Mandriva) that would assist the user into further
+system & DE configuration after the initial install makes a lot of sense.
+100 hours of developers' time  vs. 1,000,000 hours of users' wasted
+post-install time. => Develop once, use by everyone.
+
+2. The same for Gentoo-like distros (stage 1). Having everybody building
+every package for every system is useless in 99.9999% of the time, as the
+claimed "optimisations" might be of 2-3% in terms of speed, whereas the
+millions of users' wasted time (and electrical energy!) are a huge nonsense
+-- and it's anti-ecological too! => Build once, install by everyone.
+
+3. FreeBSD devs have been extremely opaque wrt binary updates. I've not been
+using FreeBSD since ages, but I believe that, even if they do have binary
+updates now, they're not in plain repositories that could be browsed
+(FTP/HTTP) like the Linux updates repos. Once again, patching the FreeBSD
+style is a huge waste of time and adds unnecessary risks.
+
+4. XFCE devs really can't understand/accept that _not_ having for the
+desktop icons a text label drawn with transparency & outline (instead of the
+default opaque background) makes their DE look like Win95. (Transparency
+came with Win98 and outlining... I'm not sure.)
+
+5. There are gazillions of _relevant_ bugs (even in applications like gedit,
+kate, whatnot) that never got fixed, whereas the developers of the
+respective DEs always try to add new features, sometimes exotic and
+irrelevant ones (compiz-included). It's like they prefer the glitter (the
+bling-bling) to _relevant_ functionality. Yet, they develop for Linux, not
+for Windows. Go figure.
+
+6. "If it ain't broken, don't fix it." Countless examples (y compris KDE4,
+GNOME3, but I'n not going into this), I'll just say SysV
+init/systemd/upstart/whatnot -- I'm not even interested in this crap. It is
+crap for me because:
+i. Booting time is irrelevant for servers, they're 99.99999% up.
+ii. Only stupid desktop users would shutdown when hybernation
+(suspend-to-disk) is available.
+iii. Gaining 10 seconds in boot time is not worthing, if the price is a
+disruptive redesign of the _entire_ init process, with tons of downstream
+work for everyone.
+
+But, as I said, I never trust the judgment of FLOSS/pro-bono developers. So
+I don't expect any of you to understand my rationale -- I even expect
+Dodonov to piss on my words, as he already did today. (Quite unexpected, as
+he already worked for Microsoft, and so far no Microsoft guy mocked me.)
+
+R-C aka beranger
+