On Sun, 21.07.13 16:09, Reindl Harald (h.reindl@xxxxxxxxxxxxx) wrote: > > Am 21.07.2013 16:05, schrieb Matthew Miller: > > On Sun, Jul 21, 2013 at 08:58:39AM +0200, Nicolas Mailhot wrote: > >> it. Yet the "something better" never materialized. When I had a disk go > >> wrong lately I was notified by the big ugly legacy system. I had *zero* > >> notification by all the "better" systems that were given as "evidence". > >> Because the "better" systems do not exist. None of the 'smtpd is legacy" > >> complainers have actually tried to solve the (remote) notification > >> problem, none of them actually understand the reliability and operational > >> constrains, or that being to define message routing (via aliases, > > > > Sure they do. I can't imagine an installation of any size (eg more than 2 > > systems) not using Nagios, Icigna, or some other alerting system. > > > > If you're in the narrow case between a desktop system and an installation > > where real monitoring and alerting is worth it, install an MTA > > the problem i see is when things like MTA and rsyslog are > removed from the defualt install many pakcgers will less > care about them in the future nor test how well it works This is not a valid argument. We cannot keep extending the core all the time by adding more and more redundant components to them, just because we believe this will cause APIs to be tested. It's the wrong approach in every way. As we all know the current sendmail setup means mails are lost in a /dev/null-like mail spool nobody reads. So we know right now that the current situation is certainly really broken. What have we done so far about it? Just ignored it, and continued to let the mail spools run over with mails that are ignored. The few people who cared didn't mind because they were the ones who reconfigured the aliases file or the entire MTA in a way that actually made it useful. But that's a pretty borked situation and unfair to everybody not in the know... You get your stuff well-tested by removing redundancy, by making your stuff interesting to people and by making sure people who care test it. It's not enough just put on a system where nobody cares about it. It didn't fix the local MTA situation in the past 10 years and it won't in the next. Also, what kind of a picture does this paint of the Fedora project anyway? Only stuff we install by default matters and is tested? If that was the case the only answer could be to drop all non-installed-by-default packages from the dirstro... - but that would just be so wrong.... Lennart -- Lennart Poettering - Red Hat, Inc. -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
