On 24/11/14 19:13, Marc Stürmer wrote: > Am 24.11.2014 um 19:25 schrieb Gevisz: > >> I switched from Ubuntu 10.04 to Gentoo just because it forced closing >> window button "x" to the upper-left corner of the window in Unity of >> Ubuntu 12.04 while I used to look for it in the upper-right corner. :) >> >> So, I see no reason that those that hate systemd would not do the same. > > I also did for my own server. > > But the real strength and home of Debian on a server is in the > corporate environment, and in a CE you are facing other challenges, > namely: > > * long term support (meaning for a few years),
I'd clarify this even more to say almost transparent to install upgrades within that release cycle > * stable releases with a more or less stable and predictable release > cycle, debatable - i would suggest "better tested" than stable otherwise there would be no need for debian bugzilla > * steady stream of security updates as long as the release is being > supported. > > Which also explains why in that field so many people are so heavily > against SystemD, because it is still: > > * quite a young software project, which needs more time to mature in > their eyes, > * still a fast moving target, with adding more features over features > with every new release, > * maybe also the philosophical aspect that it violates one of the > primary paradigms of UNIX: do one thing only and do that well, > * and it forces them to learn a new way to configure their system, if > they would use it. > +1 for all of these >> I disagree: the downloading all that crap also takes a lot of time. > > Downloading binaries takes of course some time, yes. But downloading > e.g. the source code of Chromium compared to the binary of Chromium > does take a multiply longer. And after the download of the binary you > just need to unpack it and are ready to run it, on Gentoo you need to > compile it. > I would argue the opposite. I would say that because of the portage binary features, and the "possibility an upgrade may not even compile" it forces me to do better QA on updates. as an example, I would be less likely to test and update in debian or red hat before applying a series of necessary updates. on gentoo cluster i would install off the cluster first, ensure everything went smooth then distribute the binaries. for issues with conf changes *cough ISC bind and freeradius cough* it means that i'm well prepared. it also means that continuous kernel configuration changes for the various udev updates can be masked and prepared for in a better way than "oh this week's updates require i reboot the server" good luck using custom kernel or initram with the major distros -- i found that that was a surefire method to bork things, non bootable and confused app-manager both at the same time. > So binaries are by every mean faster to download and run than > downloading the source, compiling it and then running it on a server. > Even downloading the biggest archives and installing (without > configuration) is normally done in under one minute. That's the time > saving aspect, and you got no broken ebuilds. Of course you got > another can of worms that may be bug you instead. > > And if you don't like the example of Chromium, then take MySQL e.g. > instead. > > People in a CE rarely have the time to deal with the added complicity > of Gentoo compared to binary based distributions, and therefore Gentoo > just don't fit for most of them. > +1 gentoo in a very real sense is "my distribution". my /etc and my /var/lib/portage/world and i have geezer-linux-desktop and geezer-linux-server but in a corporate environment it is someone else's problem be that low level... rather than have an inhouse developer to fix the web application bugs, they would have a "Next Generation Unified Threat Management Firewall" to block people taking advantage of those bugs. the question is how it is sold. also it is a lot easier for someone to click on the little balloon that says "updates pending" than to think about what it is they are doing. equally it is easier to convince a business to buy one server instead of trying to cluster two or more -- then you _must_ do updates at 3am, but updates are somehting that should happen when the updater is most alert imho. the business shifts the responsibility of the down time in the same way as they would shift the responsibility of the lower levels of distro management. > The thing is: compiling your own binaries on a production server is > something many people won't like, because it takes power from the > other processes away for that time. +1 > > And having a fully fledged C/C++ compiler running on your server is a > security hole, if you are paranoid enough. > +1 > Of course you could setup just a compiling server for all of your > other servers, but this takes time and adds complexity. > surprisingly little - honestly. >> Steady "release cycle" is also not so good. > > It depends on your case. > > All the major BSDs, FreeBSD, NetBSD and OpenBSD, have had a steady > release cycle - a new release every half year - for almost two decades > now and they are content with that. > OT: one day i thought to try a BSD but then as a penance for my sins i also read kuro5hin and there was a wonderfully scalding attack [1] on de raadt. the truth is probably no where near to the rant but it always think of it when i see attacks on lennart. but this did make me discover something that i thought i'd share here "BSD is a unix written by unix people for the pc Linux is a unix written by pc people for the pc" but its also interesting to note portage has a nod to BSD's "ports collection" WARNING link is not safe for work and may cause stomach ulcers [1] http://www.kuro5hin.org/story/2010/6/11/9571/98591
