On Fri, Nov 27, 2009 at 01:05:37AM +0100, Joachim Schrod wrote:
Lars Müller wrote:
On Thu, Nov 26, 2009 at 01:49:27PM +0100, Joachim Schrod wrote:
Since: the actual problem with avahi is that the documentation is crap. Just look at http://avahi.org/wiki/Avah4users and tell me with a straight face that this bunch of information is sufficient for troubleshooting
Simply change and fix it by contribution. [...] As you might have done already.
;-) No, you won't succeed luring me to it. While I know avahi &al, I don't use it in our own networks and thus won't invest work there. (I'm a member of the LaTeX team, a CTAN administrator, and active in DANTE, being a founding member; almost all OSS work that I'm doing in my free non-work time is spent in the TeX world.) Having two parallel Internet connections, using policy based routing, and running experimental services on all systems doesn't lend itself to Zeroconf -- our system configuration is highly specialized and thus controlled with Puppet. But then, I run an IT consulting company; I'm not a normal Linux user.
Nothing to add. Such a setup isn't the goal of the link local approach. And contributing to LaTeX is also a very good excuse. ;) My main goal was to express that it is much easier to contribute than most people believe. With the openSUSE project this isn't limited to the Build Service. Everyone is able to update or add a wiki page. And sometimes even the discussion at a mailing list leads to a solution and new insights. :))
PS: And while I'm at ranting, the same holds for all those new *Kit daemons, too. There is no document to be found that explains how udev, HAL, ConsoleKit, and PolicyKit are supposed to work together. Not to speak of PackageKit; just look at the recent F12 disaster... The attitude of the *Kit developers ("I don't care about the traditional Unix way"), combined with their unwillingness to provide good documentation, makes this stuff a complete mess from a sysadmin point of view.
Even this is part of the OSS concept. On the one side people like to control a WiFi NIC from the desktop while others they like to keep the "good" old netconfig approach.
I do understand the intentions behind those systems, and I agree to the sentiment that the traditional Unix way is not sufficient for modern environments with hot-pluggable components. My problem is more the way that has been chosen as a remedy, where the overall architecture of hardware management is (a) quite complex and (b) does not take sysadmin tasks in account: while the system is configurable as hell, very few people know how to connect the dots of the myriads of XML configuration files. Having mostly API documentation for systems like HAL, D-BUS, or *Kit doesn't help either. E.g., I have started a HOWTO documentation for myself, and it has still more questions than answers.
Even this would be a good starting point. Cause without the open questions we'll never see the answers. And I'm quite sure I'm not aware of 10% of the questions you'll raise. For anyone using Linux a bit longer this "new" FooKit and other stuff is different. But the lack of documentation isn't something new. Have you ever tried to understand why cron writes "permission denied" to the syslog? And it's not CronKit which is doing this. :) BTW it looks like DeviceKit it the most current kid of the kits. Which in general stresses your use of *Kit. Lars -- Lars Müller [ˈlaː(r)z ˈmʏlɐ] Samba Team SUSE Linux, Maxfeldstraße 5, 90409 Nürnberg, Germany