PB> When is a file a configuration file and when is it system-independent common
PB> data?
This is depends from program, of course.
PB> And when is it a system file and when is it a local addition?
Local addition == something created just for this comp (cluster, etc). ALL
files from distribution are not local additions.
>> /etc/X11 is the recommended location for all X11 host-specific
>> configuration. This directory is necessary to allow local control if
PB> Phooey. /var/X11/lib/xdm was alive and well on SuSE last time I looked.
PB> And on redhat too probably.
Not, of course. It's in /etc/X11/xdm as it should be according to FHS:
-- cut --
/etc/X11/xdm holds the configuration files for xdm. These are most of
the files normally found in /usr/lib/X11/xdm.
-- cut --
>> /usr is mounted read only. Files that should be in this directory
>> include Xconfig (and/or XF86Config) and Xmodmap.
PB> "X11 host specific configuration" means nothing to me as an english
PB> phrase, except for your two examples and my suggestion for xdm (except
PB> that I use kdm :-). Where do the X11 applications put their configs?
In /etc/X11 ... At least in normal
(FHS compliant) Linux systems.
PB> .. You know where some files have gone and so you say that they are
PB> "X11 host specific configuration" if they have gone in places that match
PB> that, but really, it is never absolutely clear what description files
PB> meet. I'm not even sure that XF86Config is a host specific configuration¸
PB> since I have my own (i.e. tied to my account) so that I can startx with
PB> the right options for me on whichever machine I log in at the console at
PB> (into, on , ..).
This is other story :-) Of course you could have configuration local to your
account as well...
>> PB> That's where I'd put it, but RH doesn't put it there.
>>
>> Since FHS says that it's belond to /etc/X11 ...
PB> It didn't. It recommended it.
FHS is just recomendation as whole. And you are free to do anything. This is
Linux after all! But if you want to have LSB-compatible system you should
follow FHS recomendations (FHS is basis for LSB).
>> PB> It's in something like /usr/lib/X11/etc.
>>
>> No. It's in /etc/X11 as it should be.
PB> OK - but I'm fairly sure (from helping some people by email) that xf86config
PB> on redhat puts it in /etc/XF86Config.
Yes. There are was this error more then year ago. In RedHat 5.1 xf86config
will put configuration in /etc/X11/XF86Config as it should be (just checked)...
>> PB> And what about /var/X11R6/etc?
>>
>> There are no such dir.
PB> Makes sense to me, however. There's a /var/X11R6/bin and lib.
There are NO /var/X11R6/bin, /var/X11R6/lib or /var/X11/bin or /var/X11/lib or
whatever. Why there are should be such thing as /var/X11R6/bin /var/X11R6/lib ?
>> RedHat put then in /var/lock as it should:
>> -- cut --
>> 5.5 /var/lock : Lock files
>>
>> Lock files should be stored within the /var/lock directory structure.
PB> Via symlinks? Fine, I symlink everything like /var/lock /var/locks,
PB> /var/spool/uucp to the same place. But I prefer the latter out of nostalgia.
/var/lock is THE ONLY place for lock files. There are NO /var/locks and
/var/spool/uucp is not link to /var/lock (what for? /var/spool/uucp is
directory for UUCP spool, not for locks!).
>> name of the device. For example, to lock /dev/cua0 the file "LCK..cua0"
>> would be created.
PB> And we know that is a good example :-).
:-)) Why not ? cua0 is obsoleted but still exists ...
>> This is not religion anymore. You have no choice (and this is good thing).
PB> You're not reading the specs with a lawyers eye. What does "within" mean?
I'm not interested in "lawyers eye". When some specification started to be
read with "lawyers eye" this is clear sign that this specification will not
be usefull for anything: if your goal is just a formal approval then there a
plenty ways to do something 100% compatible and 100% useless. For [almost]
any specification...
>> Different people prefer to place user accounts in a variety of places.
>> Therefore, no program should rely on this location. If you want to find
PB> And therefore where do the http logs go?
What's http logs ? httpd may be ? In /var/log or in subdirectory or /var/log ...
In case of RedHat it's subdirectory: /var/log/httpd
PB> http or www is usually a user on
most systems.
Hm. Yes. This is out of scope of FHS :-(( May be LSB will help here.
PB> Where does root live?
In / , /root or /home/root according to FHS ...
PB> Is that a user or not?
User, of course.
PB> Come to that, what
is a logfile? Is xsession-errors a log file?
PB> Then what's it doing in my
account?
This is clearly not system-wide logfile
:-)
>> out a user's home directory, you should use the getpwent(3) library
>> function rather than relying on /etc/passwd because user information may
>> be stored remotely using systems such as NIS.
PB> Absolutely. In fact, it's the only place it's stored here. But redhat have
PB> historically had the most ballsed up NIS installation imaginable. They
PB> didn't come with a ypbind, preferring to add things into the libc getpwent
PB> functions instead. I've been through getting nis to work on redhat (I have
PB> redhat on some sparcs).
Hm. Not know what's so horrible there. NIS support is integrated in glibc not
but RedHat AFAIK. And works quite well...
>> PB> fair approximation. But there really is a problem here. Placements are
>> PB> not compatible amongst the distributions. RedHat is the principal
>> PB> offender in my opinion.
>>
>> Debian with it's perl-based startup system is even worse. Since this is
PB> I like debian in general, but I've never looked at their init scripts.
PB> I don't use debian anywhere here, though I use their packages.
>> religion this aspect is not touched in FHS :-((
PB> Amen.
>> *ALL WORKS* ? This will never happens. glibc works more reliable then libc5
>> just now but you could use your libc5 systems if you wish...
PB> :-). I suppose you are saying that libc WAS unreliable, given that it hasn't
PB> changed in a while. No - it was fine before and it is fine now. New is not
PB> usually better, quite the reverse. It takes a while to get things back to as
PB> good as the way they were, and I'm still watching the bug finding rate on
PB> glibc2 with interest !
Hm. When half-year after detection EXPLOITABLE security hole there are no
new version with fix ? Is it far better then glibc2 situation ? Gosh ...
At least bugs in glibc are not left unfixed forever !
>> PB> applications. I can assure you that I make a standard set of changes in
>> PB> RH spec filezs, another set of changes in SuSE spec files, another in
>> PB> debian setups, and so on. The distribution and their libc major is the
>> PB> only key I need.
>>
>> This is not a solution at all: take a look on linuxhq and try just count
PB> Solution? I wasn't proposing a solution. It's just that i can use that
PB> information, thanks, and I'd prefer to be given it. I have to put up with
PB> config.sparc-sun-solaris2.5
PB> config.sparc-sun-solaris2.6
PB> etc. etc. and umpteen flavours of irix. I don't see why I shouldn't get
PB> config.ia32-redhat-linux5.2
PB> to help me along as well.
How many useful information you'll got from fact that I'm use KSI-Linux 2.0 ?
To me information about libc used here (2.0.7) is much more usefull...
>> distributions (including different versions! RedHat 5.x has chkconfig while
>> 4.x does not). And you want to say that all programs must be aware about all
>> distributions ??? You are on drugs :-((
PB> I didn't say so. I said that I could use the info thank you.
Not always. If you have information about libc and/or kernel -- this is always
usefull information (there are only few versions of kernel and libc :-) but
information about system name is MUCH less usefull...
>> PB> bugs in their install present since 4.0, such as not working when you
>> PB> try and make more than 9 partitions at startup, not reading results from
>> PB> fdisk, only from disk druid, not letting you change the hostname once
>> PB> chosen a first time, etc. etc.).
>>
>> Even with all glitches this is simpliest to use linux install I'm seen...
PB> Try SuSE. It's beautiful.
Hm:
-- cut --
It is forbidden to reproduce or distribute data carriers which have
been reproduced without authorisation for payment without the prior
written consent of S.u.S.E. GmbH or S.u.S.E. Linux. Distribution of
the YaST programme, its sources, whether amended or unamended in full
or in part thereof, and the works derived thereof for a charge require
the prior written consent of S.u.S.E. GmbH.
-- cut --
Have THIS as heart of "free OS" ? You are joking. I'm not crusader (like RMS)
but for me such program is clearly unappropriate and SuSE without YaST is not
very usefull (or am I wrong?). And "/var/X11/lib/xdm was alive and well on
SuSE last time I looked.
" -- this means that SuSE is not even FSSTND 1.2
compliant ! Gosh.
>> BTW I'm does not use disk druid and all works fine for me here with fdisk, etc.
PB> Not for me. It seemed to go into some kind of endles loop of requiring a linux
PB> root partition, letting me "make" it, then ignoring what I'd made. I went to
PB> the other screen and fdisked it for myself, then started disk druid in the
PB> screen and let it see what I'd done .. of course I had to mknod the right
PB> device names by myself. I think that's the catch, in fact. They probably
PB> only make the devices that correspond to the partitions seen by disk druid.
More or less. But for me all was Ok even for few systems with 5-8 partitions of
total size >20Gb...
>> Hostname could be changed easily via Control Panel (just click on Network
>> Configuration), LinuxConf (Config->Networking->Client tasks->Basic host
>> information->Host name) or via edition of /etc/sysconfig/network, etc. If
PB> You call that easy!
Why not ?
PB> What was wrong with echo foo > /etc/HOSTNAME;
hostname foo!
Why edition of /etc/HOSTNAME file is any better then edition of
/etc/sysconfig/network file ?
PB> And what is a Control Panel?
Program with name "control-panel". Started automatically for root after startx.
PB> The menu thing I was in at
startup seemed to be "linuxconf".
May be. Since there are no "menu thing at startup" (in RedHat 5.1 at least)
I'm could not confirm or decline this.
PB> There were umpteen twisty little
PB> mazes of configuration menus that in true gatesian style didn't tell you
PB> where you were going or how you could get back, but there was certainly
PB> no way of changing the hostname that i could find.
If this was really linuxconf then see above.
PB> I looked as hard as
anybody trying to get his machine to recognize its
PB> own name could
possibly try. I.e. I explored every menu.
You call THIS "hard try" ? Gosh. I'm start with find and grep in such situation
and usually in few minutes solution is found. BTW even if I'm not use not
control-panel nor linuxconf myself I found place for hostname in BOTH programs
in two minutes (while writing answer to you angry letter :-)
PB> If the menus depend on
state, then I suppose I could well have not
PB> triggered some, but apart
from that possibility, I exhausted the avenues
PB> available.
I not know much about menus in control-panel or linuxconf but 2 minutes was
enough to find this item in both...
>> you can not drive this is does not mean that you are using broken car -- may
>> be you just can not drive...
PB> I can drive fine, I assure you. My opinion is that redhat can't possibly be
PB> expected to test their vehicles out on every alleyway, and that therefore
PB> there are plenty on which it doesn't work, and no luser has ever complained
PB> because it works fine on the big wide road they drive on: a one
PB> partition, copy from cdrom, standalone machine setup that they normally
PB> use. That's to be expected. We know there are bugs in the RH config
PB> utilities .. we don't even have to look to find that out because it's
PB> deducible from their nature. I could list more bugs (if my memory holds
PB> up :-). It's just sufficiently annoying to annoy me sufficiently not to
PB> want to use their stuff.
Choice is yours. But I'm have a lot of RedHat based systems and they are mostly
works as expected -- you could change hostname in /etc/sysconfig/network and
enable/disable forwarding with the same file, you could call to ISP via
'ifup ppp0' and stop call via 'ifdown ppp0', etc. All works just fine (may be
menu-driven tools does not work fine in RedHat -- I'm simple not know since
I'm not use them :-)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/