Il 01/12/2014 12:38, Arrigo Marchiori ha scritto:
On Mon, Dec 01, 2014 at 06:31:18AM +0100, Alessio 'Blaster' Biancalana wrote:No ma parliamone. :D A voi systemd fa tanto schifo? :PSì, IMHO. :-(
 Un esempio illuminante, che non avevo letto prima. Grazie Arrigo.
Sarà un caso, ma ieri ho letto un articolo su Linkedin di un ingegnere software che elenca quelli che ai suoi occhi sono i grossi problemi che ha causato, sta causando e, soprattutto, minaccia di causare nel futuro systemd, di natura non solo tecnica, ma anche politica (nell'ambito dello sviluppo di Linux). Ve lo riporto in lingua originale:
Bill "TTK"
Software Engineer at Sonic.net
If it were just systemd vs SysV init, I suspect people would be less upset. Unfortunately systemd poses several drawbacks, which taken as a whole make me want to avoid the entire thing.
For instance, systemd replaces a great deal of well-understood mature software with poorly-understood new (and therefore buggy and vulnerable) software:
* init
* network configuration
* DNS resolution
* iptables
* ntp
* ConsoleKit
* dhcp client
* cron
* atd
* pm-utils
* inetd
* acpid
* watchdog
.. and others
Also, in order to function as an init system (sans its temporary sysV compatibility mode), it requires invasive hooks into each service's source code, for what they're calling "socket activation".
Furthermore, it owes its widespread adoption NOT to its technical merits, but to deliberate manipulation of software dependencies. systemd was mostly *ignored* until udev (which is how most distributions load device drivers) introduced a systemd dependency.
Once this happened, most of the major distributions adopted systemd to continue using the latest udev releases.
Not all, though. Gentoo forked udev to remove the dependency (eudev), and Slackware and Debian use old, pre-systemd versions of udev (though political machinations in Debian's technical committee have introduced systemd to Debian anyway, causing considerable strife within the Debian community, and ultimately this fork).
After eudev was picked up by several other Linux distributions MailScanner ha rilevato un possibile tentativo di frode proveniente da "www.linkedin.com" http://forums.gentoo.org/viewtopic-p-7648392.html Lennart declared the intention to have device driver developers rewrite their drivers to use the systemd-dependent kdbus interface instead of the traditional netlink.
If this is successful, it will increase the maintenance burden of using eudev (as must-have new device drivers would have to be ported to use netlink).
On top of this, the systemd developers are outspoken about bringing all distributions "into the fold", arrogant and derogatory about people's efforts to avoid this (like XFCE's development of ConsoleKit2, to avoid a systemd-logind dependency), and disdainful about fixing reported bugs.
Their rhetoric is simple and effective, but also wrong. They both implicitly and explicitly state that universal adoption of systemd is inevitable, when it is anything but. They say that nobody is offering an alternative to systemd, when there are many alternatives. They say systemd opponents are just resistant to change, pretending that merit-based arguments are never posed. This is the kind of rhetoric used to win for winning's sake.
The systemd developers are largely ignorant of the hard-earned lessons which have contributed to the security and robustness of the userland components it is replacing, with dire consequences. See, for example: MailScanner ha rilevato un possibile tentativo di frode proveniente da "www.linkedin.com" http://www.openwall.com/lists/oss-security/2014/11/12/5
The software systemd replaces represent a HUGE attack surface. The DNS exploits enumerated in the above link is just the tip of the iceberg.
The obvious solution is to simply not use systemd, and continue using one of the distributions which shun it, but there is a danger in doing so. Because of the narrative that EVERY distribution has adopted or will adopt systemd (except for a few hangers-on who don't matter), it is possible that software developers will start assuming that every Linux target will have systemd, and introduce a hard dependency on it. If this happens, distribution developers will face the burden of excising these hard dependencies from must-have packages.
To prevent this from coming about, the narrative of systemd ubiquity needs to be countered with a strong demand for support for non-systemd Linux targets. As long as the must-have packages provide a --disable-systemd configure option (or similar), using a systemd-free Linux distribution will continue to be a viable choice.
Devuan is a big step forward in making that demand visible. I wholeheartedly wish it the greatest success!
-- Alessandro Selli " target="_blank">< > Tel. portatile: 340.839.73.05 VOIP SIP: " target="_blank"> Sito web: http://alessandro.route-add.net/ Chiave firma PGP/GPG signing key: B7FD89FD
Archivio con motore MhonArc 2.6.16.