sm-discuss AT lists.ibiblio.org
Subject: Public SourceMage Discussion List
List archive
- From: Seth Alan Woolley <seth AT positivism.org>
- To: Casey Harkins <charkins AT upl.cs.wisc.edu>
- Cc: sm-discuss AT lists.ibiblio.org, Laurent Wandrebeck <low AT mren2.univ-littoral.fr>
- Subject: Re: [SM-Discuss] 0.8.0-pre2 iso
- Date: Wed, 10 Dec 2003 20:36:16 -0800
On Wed, Dec 10, 2003 at 03:08:37PM -0600, Casey Harkins wrote:
>
> On Wed, 10 Dec 2003, Laurent Wandrebeck wrote:
> > sysklogd is imho necessary just in case something goes wrong...how would
> > you do if, a apparently correctly configured device doesn't work without
> > any logger ? suppose it's your network device, you can't browse the web to
> > install a logger...good luck. that's one the reasons for the logger
> > presence. a second one would be a bit of security. a minimal system
> > doesn't mean to me a blind one.
>
> If someone has the knowledge to check the logs, they should be able to run
> the network tools themselves with verbose output. Or perhaps we could add
> a -v (verbose) switch to the ifup/ifdown scripts.
>
> I understand your desire to have a logger by default, but forcing a
> particular logger on a user, even if it is only until they can dispel and
> cast their preferred logger is a bad idea. It'll leave the /var/log
> directory cluttered with sysklogd's logs if the new logger uses different
> log filenames or directory structure (ie metalog). If it is decided a
> logger *must* be installed, each of the loggers source/dependencies could
> be included in /var/spool/sorcery in image.tar.bz2 and the installer
> could prompt for one to cast.
>
> -casey
logger isn't needed for the new init.d anymore. The main reason I added
it was because I tested the new init system on a sometimes monitorless
(headless) server (I have a 600 ft^2 apt. with a wife and very little
room for clutter). Since the init.d system had to refactor echos in the
scripts to get them to parallize by line and not print over the top of
one another, logger support was made _really_ easy since it too expected
whole lines of data being sent. So when debugging spells, I used the
init.d logging to help track down issues without having to have a clunky
monitor in the corner of my almost "studio" apartment (it does have a
separate bedroom where no computers are allowed, though). Logging is
now optional and disabled by default, as well, now.
Of spells that install init.d scripts, logging is entirely optional
(network, for example, checks a USE_SYSLOG variable). So, the logger
could conceivably be dispelled and a reboot would work fine in a default
install. (I haven't tried this myself, but it works in theory. ;) )
As far as security is concerned, logging _is_ important, and having a
logger on each iso really helps us in the irc chat room debug problems
that may arise. But most people planning on installing Source Mage in
pre-existing environments will have their own preferred logger. Yes,
sysklogd can be dispelled, and another installed in its place, but 1)
those who need help without a logger in the chat room can simply cast
one and try again, and 2) most places with logger needs will have
selected their favorite logger to configure anyways. I say it is
entirely possible to leave the logger off in the default install without
much harm done. Any user not a pre-existing user or coming to the chat
room will either know to install a logger when they want one, or if they
don't, won't know what to do with one anyways.
As far as other spells included in the iso, if somebody needs them to
install a system, leave them in, please. I really don't want to have to
install debian so that I can install Source Mage, just because I have
awkward first-boot requirements. We can remove the truly first-boot
unneeded stuff like xdelta and sysklogd.
Why do I think xdelta is bad? It's relatively unmaintained, and it has
no bzip2 support. No bzip2 support is the killer for me. If we do use
xdelta, we should ignore the built-in tar.gz support and do the delta on
.tars (the delta itself would be gzipped, of course, for transit). I
was going to consider just doing diff/patches, but since most version
bumps rename the main directory the files are in, diff -Nuar on the
directories would be ... useless, unless we could work around that with
custom coding. Getting useful bzip2 support put directly into xdelta
would be really nice, too, but I think it's non-trivial (but a better C
coder than I could certainly take a shot at it).
Just my thoughts on the matter...
Seth
--
Seth Alan Woolley <seth at positivism.org>, SPAM/UCE is unauthorized
Key id EF10E21A = 36AD 8A92 8499 8439 E6A8 3724 D437 AF5D EF10 E21A
http://smgl.positivism.org:11371/pks/lookup?op=get&search=0xEF10E21A
Attachment:
pgpQBhBDKGswM.pgp
Description: PGP signature
-
[SM-Discuss] 0.8.0-pre2 iso,
Andrew, 12/09/2003
- Re: [SM-Discuss] 0.8.0-pre2 iso, Hamish Greig, 12/09/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Laurent Wandrebeck, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Andrew, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Laurent Wandrebeck, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Casey Harkins, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Seth Alan Woolley, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Arwed von Merkatz, 12/11/2003
- Re: [SM-Discuss] 0.8.0-pre2 iso, Arwed von Merkatz, 12/11/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Arwed von Merkatz, 12/11/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Seth Alan Woolley, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Casey Harkins, 12/10/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Laurent Wandrebeck, 12/10/2003
- Re: [SM-Discuss] 0.8.0-pre2 iso, Hamish Greig, 12/11/2003
-
Re: [SM-Discuss] 0.8.0-pre2 iso,
Andrew, 12/10/2003
Archive powered by MHonArc 2.6.24.