Re: next debian stable ?
David Krider said:
> How about: BECAUSE THAT'S THE LAST TIME THE DISTRIBUTION WAS ANYWHERE
> CLOSE TO BEING CURRENT. And it doesn't look like it's going to change any
> time soon. (And I'm yelling, not at you, but in frustration of that
> fact.)
I hope it doesn't personally :) you mention below that you use a lot
of solaris and EMC stuff. I would expect you more then others to
appreciate runinng tried and tested software. I mean theres still a lot
of folk out there that use solaris 2.6, or HPUX 10.20.
> charity job. Off the top of my head, I need to install FrontPage
> extensions at at least two of these 3 types of locations. If you know
> anything about this miserable product, you know how difficult it is to
> install in Apache 1.3 servers.
yeah I don't miss frontpage :) I haven't touched it in nearly 4
years. it was a pain to get it all working, but last I remember
"ready to run" software had decent instructions at the time..
when I did install it(debian 2.0 at the time), I installed it
completely from source, statically linked all the apache modules,
I think it installed to /usr/local/apache or something.
> install it from unstable." I haven't been through this exercise
> specifically, but I'm willing to bet there are a lot of dependencies
> there.
if you are running woody, do NOT apt-get install ANYTHING from
unstable. NEVER EVER EVER. you will likely hose your system, or
upgrade it to unstable. Though building from source from unstable
may be possible:
(assuming the package name is apache2, haven't looked myself):
1) add source uris to /etc/apt/sources.list (there should already
be some, the lines start with deb-src, change references on those
lines from stable to unstable
2) apt-get update
3) apt-get build-dep apache2
(hopefully you can get all the deps from woody otherwise you'll
need to build from source those deps that you cannot)
4) apt-get source -b apache2
(hopefully it builds successfully)
if it were me I'd stick to apache 1.3.x. though like I said I haven't
used frontpage in nearly 4 years so maybe it's gone downhill since.
or you could build it just like you would on a solaris system, download
from www.apache.org/httpd/ and go from there :)
> Where do I learn how to do this? Why, by living in such a mess on
> the desktop, of course. Trying to pull in just, say, a better kernel, is
> a perfect example of how this process is supposed to go in the general
> sense. That's what I mean about "mastering." The process isn't nearly as
> clean as fans of apt would like to make everyone else think it is.
as for kernels I've been building mine from souce ever since my
slackware days in 1996, out of habbit mostly but it turns out at least
it's been good, e.g. the recent linux ptrace local root bug didn't
affect my systems since I don't compile CONFIG_KMOD into my kernels
(and my kernels are 2.2.x). I have yet to put a 2.4.x system into full
production, most of my important servers are 2.2.19 still(my desktop
is too, though my laptop is 2.4.20, using the laptop as a testbed for
2.4.x to see if I have any troubles, sofar none).
> 1) How many times have you "seamlessly upgraded" your Debian boxes to a
> new version? Given the historical release schedule of the project, it
> CAN'T have been more than once. For all the hassle I've had with
> learning the structure of the Debian system, I could have paid for fresh
> installs or SuSE or Red Hat several times over.
I know your not asking me but I'll answer anyways, I've upgraded
about 35 systems pretty seamlessly. minor glitch here and there
sometimes but nothing serious. Biggest problem was one of my firewall
boxes had some "iffy" memory in it, apt-get segfaulted during an
upgrade from 2.2->3.0. ack! I emailed this list but no response,
I had it fixed in about 20 minutes had to do some dirty work with
dpkg with --force-depends and --force-overwrite but I got the
system back into a consistant state and was able to complete the
upgrade. Another system a co-worker upgraded about 4 months after
I was laid off, I warned him ahead of time of potential problems
such as the sendmail I had built probably would break HARD in
the upgrade, so told him to migrate to postfix, easier. I didn't
anticipate cyrus breaking in the upgrade though, not sure why
it did, but a quick dpkg -i manually on the cyrus packages fixed
the problem(all of this done over the phone, I didn't need to
login to the box). I think i even upgraded a few systems
from 2.1 to 2.2 back in the day. but have done TONS of 2.2 to
3.0 upgrades.
contrast:
FreeBSD 4.4->4.6 upgrade(last year). "make world" and stuff goes
smooth, I decide to wait until later to reboot the box with the
new kernel. To my suprise! ipfw is no longer functional, it
segfaults immediately when run. Damnit! Have to reboot now to
get it working, so I reboot, to my suprise! /etc/sysctl.conf
format has changed(ever so slightly), so my bridging configuration
breaks, and now both of my networks are down!(everything goes
through that machine), so I bypass the bridges and work on it
till I discover the update. Oh and i wanted to wait to reboot
the system since I was using a binary-only driver for a 4
port network card I had which I had no way to test to see if
it would work in 4.6(though I suspected it would, and it did),
since it was built for freebsd 4.0.
FreeBSD 4.7-4.8 upgrade(bout a month ago). "make world" and
stuff goes fine, I want to wait until later to reboot with
new kernel. but to my suprise 'ps' is no longer functional.
have to reboot now. luckily the system was otherwise fine.
I personally hate rebooting linux/unix/bsd machines. it always
scares me(despite most reboots going smoothly).
SuSE 7.3->8.0 upgrade(last year). on my sister's computer, a dual
P2-233 w/384MB ram and 3xSCSI drives(at the time). The install
goes flawlessly, reboot and the system immediately freezes. no
errors on the screen, the machine locks before the fancy framebuffer
can initialize. about an hour later I get to reconfiguring the
boot loader so it doesn't go to framebuffer mode and see it
dieing on the ACPI code. 2 hours later I finish compiling a new
kernel(the machine is slow and SuSE's kernels are HUGE) without
ACPI and the system boots great. I could of been up faster if
I had just installed the uniprocessor kernel but I wanted to
make use of the 2nd cpu. oh and all kernel boot parameters
to disable acpi failed, the system still froze.
and after running SuSE I realize how great I have it with
apt-get:
SuSE 8.0 - fresh install. login as root, first thing I wanna do
is run YOU(Yast online update) and get the latest packages. So
I do, YOU reports a GPG error on every package, I get tired
of clicking "OK" for each package and try to hit abort, nothing.
So I kill it gracefully (I think I just used -TERM). Fire up
YaST again, fire up YoU ..segfault. Try again..segfault. try
loading it from the console YaST(vs X11), segfault. Use YaST
to remove/purge all YoU components, then reinstall them,
still segfault. After about an hour of troubleshooting I decide
fuckit and reinstall(it was a fresh install, faster then
troubleshooting it further). turns out it was a known bug.
SuSE 8.1 - fresh install. login as root. first thing I wanna
do is run YOU, and get the latest packages(sound famillar?).
So I do, and YOU reports cannot find any files on server, try
another server(something along those lines). So I do, same thing,
I try all of their servers, same error. I use YaST to configure
a proxy server so it can proxy through my local squid, same
error. I try later that night on another machine, same error.
I remembered the first incident with YOU and checked their
support site. Yep, another known bug, YOU doesn't configure
the system for passive ftp, so I have to edit /etc/wgetrc
and set passive ftp. Then it works like a charm.
SuSE 7.2(?)/SPARC - fresh install. login as root first thing
I wanna do is run YOU and get the latest packages(sound famillar?).
So I do, and YOU downloads a crapload of stuff, but then reports
package X cannot be found. So I go to another mirror and it
grabs some more packages, then it says package Y cannot be found,
so I go to another mirror and it finds package Y but then cannot
find package Z. So I try another mirror, and no package Z. try
the rest of the mirrors and theres no trace of package Z. Even
on their main site. Why they would put something on their site
that says package Z needs to be updated but then not provide
package Z anywhere is unusual. but to be fair I don't think their
sparc port is officially supported.
that said I do love suse, i think it has it's strengths, I
reccomend it to linux newbies, my mom has it on her laptop,
I put it on my sister's computer, I bought 2 copies of 7.3
(ironically YoU worked flawlessly in 7.3, the first version
it came out on), 2 copies of 8.0, 2 copies of SuSE Pro Office
(packaged version of staroffice 6), and 1 copy of 8.1, haven't
bought 8.2 yet. My sister asked me how she could use Yahoo!
messenger, I fought with "everybuddy" for a bit trying to figure
out what firewall rules I needed and stuff, and told her I'd
get back to it later. she had downloaded the win32 version of
Yahoo! and i told her it would not work, not compadible. but
she tries it anyways. and it works! installed and runs with
wine, no help from me. not only was I impressed it worked
at all(thought wine would need some config at least), but
more so that my sister could do it without help, I mean she
barely knows anything about computers.
> 3) Maybe rpm-based systems don't respect admin changes _as well as_ apt,
> but before major changes, you should have a backup of your config anyway.
> My personal taste is to let it overwrite my changes with a fresh file,
> then merge my changes back in. I've been doing that for quite some time,
> and I was doing it on Debian too.
one of debian's biggest strengths I think is live upgrading though,
if you wanna upgrade SuSE(as far as I know) the only officially
supported way is to reboot and start the upgrade from the CD/bootdisk.
same for redhat. FreeBSD you can make world and make installworld but
if something changed that breaks ps well you may need to reboot. apt-get
doesn't touch the kernel during upgrades, and I've never had to reboot
a debian box after an upgrade. when my former co worker upgraded that
mail server while I was on the phone he said what a shame to lose
this 6 month uptime.. I said, what are you talkin about, you don't
have to reboot!
> You're joking, aren't you? There's no way a corporation will come
> anywhere within miles of using Debian for heavy lifting. Support is the
> big issue here. Unless you go get Progeny to build a custom Debian distro
> and subsequently support it, you're hosed. On top of that, most
> "corporate" type software, like big RDBMS and CAE software (both of which
> we are using) have very specific support requirements, and guess what?
> Debian stable won't fulfill them in many cases.
this is true.. though for many smaller companies that don't run
such software(my last employer was one), that is a non-issue. when
I started most everything was redhat, when i finished most everything
was debian(and the users that managed the network before I was there
liked it, they only used redhat at the time since that's all they
knew).
> I never thought anyone on this list would be. In fact, I'm surprised I
> haven't heard, "Don't let the door hit you on the way out" yet. ;-)
I haven't said that for AT LEAST 2 weeks.
nate
(used solaris, aix, digital unix, hpux, irix, freebsd, openbsd, debian,
slackware, redhat, suse, stampede, corel, caldera, turbolinux, and
probably a couple others I can't think of, prefer debian over all of
them any day of the week - just a personal preference.)
Reply to: