Re: What do you wish for in an package manager?

[Joseph Carter]

On Sun, Dec 24, 2000 at 07:54:00PM -0900, Ethan Benson wrote:
> personally the plain text database is one of dpkg's greatest assets.
> its a royal pain to repair a binary database when it gets fscked.  and
> yes i have already been saved from a total reinstall through the
> ability to fix dpkg's broken database with a text editor.
> 
> if your talking about a different database then nevermind.  

Berkeley DB to text and back is pretty easy.

Also, I was speaking of binary indices which are even easier to regenerate
if corrupted.

-- 
Joseph Carter <[EMAIL PROTECTED]>Free software developer

 these stupid head hunters want resumes in ms word format
 can you write shit in tex and convert it to word?
 \converttoword{shit}



pgpHnGogir7LX.pgp
Description: PGP signature

Re: aacraid driver

[NOKUBI Takatsugu]

In article <[EMAIL PROTECTED]>
[EMAIL PROTECTED] writes:

>> I'm the developer of the aacraid driver. I've noticed some old posts on your
>> web site w/r/t this driver. If the maintainer of the debian kernel is
>> including this driver, he/she should contact me at [EMAIL PROTECTED]
>> to make sure they have the latest version. 

It seems that the boot image of debian2.2r2 has not aacraid driver.

>> If your still using the version
>> snagged off of Red Hat's 2.2.16-22 kernel SRPMS, there has been a patch
>> since then that fixes a memory leak. I would be more than happy to provide
>> this patch to debian for posting on their web site or ftp site. Please let
>> me know.

Hmm, do you have the official web site of aacraid driver?
I had been replied from Adam Di Calro
, however I
couldn't find the official distribution site of aacraid driver. So I'v
not to be able to reply to Adam.

Recently, I had met some trouble on aacraid 1.0.3, so I changed to
1.0.6 (snaged off of rawhide's kernel-2.2.17-8.src.rpm) and it makes
to be better.

If official installer had it, I'm very happy (and maybe many people are).
-- 
NOKUBI Takatsugu
E-mail: [EMAIL PROTECTED]
[EMAIL PROTECTED] / [EMAIL PROTECTED]

Need to clone machines efficiently - help?

[Erik Winn]

Hi Folks,

 I have just started working with a group here in Portland that is taking in 
old machines and recycling them - putting linux on as the OS (of course ;}). 
See http://www.freegeek.org for more. Its a non-profit all volunteer thing; 
and actually one of the people has posted to one of these lists before ... 
but, apparently things kinda (lamely, IMO) drifted toward a mandrake system 
-- I think I can turn that around with a little help; I am putting in some 
good time there and I think that when the realities of maintaining and 
upgrading rpm systems hits they may change their minds.

 Here is the first obstacle - not really a big one, but I spent all day 
digging around and couldn't really find any tools for this one: we want to be 
able to clone the machines easily over the local net. Mandrake has a tricky 
boot floppy that asks only for the eth0 config and then runs a bunch of perl 
to do the rest of the install non-interactively. I haven't started reading 
the scripts yet (that's plan B), instead I was hoping that someone had come 
up with something similar for debian. We are looking at hundreds of boxes 
already and its really just begun.

 This is really not a huge task - it would just make a nice splash over here 
if we could come up with something ...

 I would greatly appreciate recieving help/ideas/advice on this - Note 
however that I am not actually subscribed to the list at present (sorry, just 
too much at the moment ) so you can reply to me personally if you like.

 Thanks very much in advance! I hope we can get this to happen.

Erik Winn

Re: Need to clone machines efficiently - help?

[Aaron Lehmann]

On Mon, Dec 25, 2000 at 12:15:50AM -0800, Erik Winn wrote:
>  Here is the first obstacle - not really a big one, but I spent all day 
> digging around and couldn't really find any tools for this one: we want to be 
> able to clone the machines easily over the local net.
> boot floppy that asks only for the eth0 config and t

While I was at VA, I worked with SystemImager, which is available at
http://systemimager.org. It might be what you're looking for. It
requires the hardware to be basically identical across machines. Once
you set up the server, which is a significant task, cloning the master
client onto the other clients is a cinch, involving putting the
automatically-generated floppy disk into the disk drive and turning on
the computer.

I wrote a setup guide for SystemImager a few months ago.
Unfortunately, I seem to have lost it :-(.


pgpHhfTcfjTw0.pgp
Description: PGP signature

How Do You Create Leads?????

[kneasi]

If you have a product or service that requires lead generation, you will
want to know about our services! 
 
We are a professional commercial e-mail service that has been in business
for over 6 years. We can help you reach more potential customers through
e-mail.  No doubt you've heard every horror story about UCE, but frankly
it's not true.  There is a very small element that does not want the
Internet to be used for commercial purposes, but isn't that what the
Internet really is?  And it completely legal to send commercial e-mail.
 
We do honor remove requests from people that do not want to receive
commercial e-mail.  In fact our remove list is over 6 million.  By the way,
our data base is 60 million strong.  If you want target e-mailing we can do
that too!
 
We've mailed for hundreds of Fortune 500 companies over the last 6 years,
and many of them you would recognize.
 
Here are some of the advantages:
 
No Printing Cost!
No Handling (stuffing envelops)!
No Postage!
Cost Effective Too!
 
We are looking for just 3 new customers  
 
For complete details, phone 248.736.3364

Re: looking for replacement for run (because of critical bug in

[Marc Haber]

On Sat, 23 Dec 2000 15:41:54 -0500, Matt Zimmerman <[EMAIL PROTECTED]>
wrote:
>On Sat, Dec 23, 2000 at 06:46:46PM +, Marc Haber wrote:
>> You'd have to have a ton of precautions. The task at hand seems
>> trivial, but it isn't :-(
>
>init does a good job of this; if there were an easy, error-proof way to add
>entries to inittab (i.e., without editing the file in your maintainer scripts),
>using init's 'respawn' mode might not be a bad idea.

Yes, but I'd want to have $daemon stoppable by init.d scripts.

Greetings
Marc

-- 
-- !! No courtesy copies, please !! -
Marc Haber  |   " Questions are the | Mailadresse im Header
Karlsruhe, Germany  | Beginning of Wisdom " | Fon: *49 721 966 32 15
Nordisch by Nature  | Lt. Worf, TNG "Rightful Heir" | Fax: *49 721 966 31 29

Re: What do you wish for in an package manager?

[Petr Èech]

On Sat, Dec 23, 2000 at 10:47:00PM -0600, Dwayne C . Litzenberger wrote:
> Hello!
> 
> I'm starting work on a new linux package manager.  The idea is to be able to
> replace rpm, dpkg, apt, dselect (backend) with one,written mostly from scratch
> and designed to be as simple (code, not features) and clean as possible.  For
> now, the work will be strictly academic, but if it works out, it may evolve
> into future standard package manager.
> 
> So my question is: What do you wish for in a package manager?

dpkg + something to handle package splits/merges a bit more sanely.
netbase split to XXX packages

php4-cgi-gd and php4-gd merges. How to handle upgrade of php4-cgi-gd to
php4-gd?

or some kind of a "dummy" package, which installes and after install it
auto-magicaly dissappears.

Petr Cech

Re: looking for replacement for run (because of critical bug in

[Marc Haber]

On Sat, 23 Dec 2000 15:41:54 -0500, Matt Zimmerman <[EMAIL PROTECTED]>
wrote:
>However, I would say that if the program dies so frequently that it needs a
>wrapper like this, it should probably be fixed.

console-log uses less syslog which dies every time the user types "Q".
And it needs to die if the log has been rotated.

Greetings
Marc

-- 
-- !! No courtesy copies, please !! -
Marc Haber  |   " Questions are the | Mailadresse im Header
Karlsruhe, Germany  | Beginning of Wisdom " | Fon: *49 721 966 32 15
Nordisch by Nature  | Lt. Worf, TNG "Rightful Heir" | Fax: *49 721 966 31 29

Re: What do you wish for in an package manager?

[Tollef Fog Heen]

* Ethan Benson 

| personally the plain text database is one of dpkg's greatest assets.
| its a royal pain to repair a binary database when it gets fscked.  and
| yes i have already been saved from a total reinstall through the
| ability to fix dpkg's broken database with a text editor.

semi-intelligent caching where you check if the text file is newer
than the binary file, if so, you use the text file and write a binary
copy to the binary file.  Else, you just load the binary.  Something
screws up?  Delete the binary file.  Make a conffile somewhere so one
can choose to use just text, if one has serious size constraints.

Saves speed, not very hard to implement (without me having looked at
dpkg's code).

-- 

Tollef Fog Heen
Unix _IS_ user friendly... It's just selective about who its friends are.

Re: looking for replacement for run (because of critical bug in

[Matt Zimmerman]

On Mon, Dec 25, 2000 at 01:29:44PM +, Marc Haber wrote:

> On Sat, 23 Dec 2000 15:41:54 -0500, Matt Zimmerman <[EMAIL PROTECTED]>
> wrote:
> >However, I would say that if the program dies so frequently that it needs a
> >wrapper like this, it should probably be fixed.
> 
> console-log uses less syslog which dies every time the user types "Q".
> And it needs to die if the log has been rotated.

It would be nice if less included a feature to close and reopen the current
file.  Then this would not be necessary.

Though maybe this program is forcing itself on the wrong interface to syslog.
Maybe it should really be reading from a fifo, a la /dev/xconsole, rather than
trying to read from files.

-- 
 - mdz

hwtools starting script

[Matus \"fantomas\" Uhlar]

Hello,

when looking into /etc/rcS.d (it's the first directory with RC scripts that
are processed at boot, isn't it?) I found this order of scripts:

README
S05keymaps-lct.sh
S10checkroot.sh
S20modutils
S30checkfs.sh
S30procps.sh
S30setserial
S35devpts.sh
S35mountall.sh
S40hostname.sh
S40networking
S41rdate
S42portmap
S45mountnfs.sh
S48console-screen.sh
S50hwclock.sh
S55bootmisc.sh
S55urandom
S60hwtools

I really wonder why hwtools is the last one in this directory. It contains
hdparm command that can speed up hard-disk access and shorten the boot time.
Why isn't this the first file? 
-- 
 Matus "fantomas" Uhlar, sysadmin at NEXTRA, Slovakia; IRCNET admin of *.sk
 [EMAIL PROTECTED] ; http://www.fantomas.sk/ ; http://www.nextra.sk/
 You have the right to remain silent. Anything you say will be misquoted,
 then used against you. 

---
Odchozí zpráva obsahuje viry.
Zkontrolováno antivirovým systémem AVG (http://www.grisoft.cz).
Verze: 6.0.217 / Virová báze: 102 - datum vydání: 1/12/2000

Re: What do you wish for in an package manager?

[Eray Ozkural \(exa\)]

Joseph Carter wrote:
> 
> On Sun, Dec 24, 2000 at 07:54:00PM -0900, Ethan Benson wrote:
> > personally the plain text database is one of dpkg's greatest assets.
> > its a royal pain to repair a binary database when it gets fscked.  and
> > yes i have already been saved from a total reinstall through the
> > ability to fix dpkg's broken database with a text editor.
> >
> > if your talking about a different database then nevermind.
> 
> Berkeley DB to text and back is pretty easy.
> 
> Also, I was speaking of binary indices which are even easier to regenerate
> if corrupted.

When I talked about binaries, people flamed me ruthlessly and without
legitimate reasons. Hope your comments make them grok the point.

Having bulky stupid slow text files all around your system is no 
guarantee of "recoverability" or "reliability", whatever. Such feature
comes from smart coding, not that things are text.

The ideal system would be that doesn't need your "manual" intervention
to keep things running. Example: checkpointing of package information
which we _don't_ have. Some random postinstall file gets fscked and
the system stalls, then you gotta fix it by hand. Ah, and if a large
portion of the database is gone, there is nothing your stupid hand
can do, anyway.

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: What do you wish for in an package manager?

[Jeffry Smith]

"Dwayne C . Litzenberger" <[EMAIL PROTECTED]> said:

>-- On Sun, Dec 24, 2000 at 11:44:13AM -0500, Adam Lazur wrote:
>> Dwayne C . Litzenberger ([EMAIL PROTECTED]) said:
>> > So my question is: What do you wish for in a package
manager?
>> 
>> Relocatable packages so a user can do an individual package
install into ~
>> without being r00t (this may be possible now with some dpkg
foo?).
>> 
>> The ability to install more than one version of a package
simultaneously.
>
>Hmm.  That could bring about some problems.  It would require a
large
>re-structuring of the filesystem hierarchy.  Why would you need
this; how
important is this to you?


Don't see why it would require a restructuring of the filesystem
hierarchy.  All that I read that as is allowing a user to install
binaries into his home directory as himself, and have the package
manager know how to find the root db, as well as his.  This would be
very nice to have, as it would allow normal users to run apt / dpkg /
whatever to install things that they run as themselves, and not have
to have the root password.  One of the strengths of Unix is this
ability, but the current packaging doesn't support it (as far as I
know).

jeff smith
-
thought for the day:  If a guru falls in the forest with no one to
hear him, was he really a
guru at all?
-- Strange de Jim, "The Metasexuals"

Re: ITO: zope-popyda, python-popy

[Wichert Akkerman]

Previously Federico Di Gregorio wrote:
>   after being asked so by the upstream author i declare my intention
> to orphan (from now) the following packages:

Why does upstream want you to orphan them?

Wichert.

-- 
  _
 /   Nothing is fool-proof to a sufficiently talented fool \
| [EMAIL PROTECTED]  http://www.liacs.nl/~wichert/ |
| 1024D/2FA3BC2D 576E 100B 518D 2F16 36B0  2805 3CB8 9250 2FA3 BC2D |

Re: ITO: zope-popyda, python-popy

[Federico Di Gregorio]

Scavenging the mail folder uncovered Wichert Akkerman's letter:
> Previously Federico Di Gregorio wrote:
> > after being asked so by the upstream author i declare my intention
> > to orphan (from now) the following packages:
> 
> Why does upstream want you to orphan them?

he developed popy and zpopyda while working for my company. he didn't leave
in very good terms. some week ago he deleted the sources from our cvs and
then told me he wanted to move to sourceforge. we said yes and gave him the
copyright over the parts written by us because we didn't want to go against
him.

i think he asked me to orphan his programs because he thinks i am a little
bit behind the upstream releases not because i have a lot of work but
because i want to damage him (by giving him bad reputation not releasing
.debs...) he also told me that some french debian developers were very
displeased with me (he didn't give any name.)

i think it is much better for me just to orphan the packages as he asked.
we (as a company) already have enough problems with him and i din't wanted
to add debian to out arguments.

ciao,
federico

-- 
Federico Di Gregorio
MIXAD LIVE Chief of Research & Technology  [EMAIL PROTECTED]
Debian GNU/Linux Developer & Italian Press Contact[EMAIL PROTECTED]
   Don't dream it. Be it. -- Dr. Frank'n'further

Re: What do you wish for in an package manager?

[Mark Seaborn]

"Dwayne C . Litzenberger" <[EMAIL PROTECTED]> writes:

> So my question is: What do you wish for in a package manager?

I want a system where I can install multiple versions of a library (or
any package really) and say which version I want each program on the
system to use, possibly on a per-user basis.  The present system is a
disaster waiting to happen:  If I install a package from unstable, it
often wants to replace my version of libc from stable with one from
unstable.  If this new libc is broken it could bring down the whole
system, when what I really want to happen is for the new package to
use the unstable libc and everything else carry on using the stable
libc.

I want the package system to behave more like a module system for a
programming language, particularly a parametric module system such as
PLT units.  I've put a few e-mails on this topic at
, if you want to have a
look.

-- 
 Mark Seaborn
   - [EMAIL PROTECTED] - http://members.xoom.com/mseaborn/ -

  ``Jersey's Crime Prevention Team are out and about,
  so have you locked up your property?'' -- Roger Bara

Re: Need to clone machines efficiently - help?

[Erik Winn]

Hi Aaron,

 Thanks very much for the pointer - I'm reading the docs for it and it looks 
very promising. Might even be worth building a couple of debs from it ... no 
promises on that right now though :).

 Happy "whicheveryouprefer"!

Erik Winn

On Monday 25 December 2000 01:08, Aaron Lehmann wrote:

> > On Mon, Dec 25, 2000 at 12:15:50AM -0800, Erik Winn wrote:
> >  Here is the first obstacle - not really a big one, but I spent all day
> > digging around and couldn't really find any tools for this one: we want
> > to be able to clone the machines easily over the local net.
> > boot floppy that asks only for the eth0 config and t
>
> While I was at VA, I worked with SystemImager, which is available at
> http://systemimager.org. It might be what you're looking for. It
> requires the hardware to be basically identical across machines. Once
> you set up the server, which is a significant task, cloning the master
> client onto the other clients is a cinch, involving putting the
> automatically-generated floppy disk into the disk drive and turning on
> the computer.
>
> I wrote a setup guide for SystemImager a few months ago.
> Unfortunately, I seem to have lost it :-(.


Content-Type: application/pgp-signature; charset="us-ascii"; 
name="Attachment: 1"
Content-Transfer-Encoding: 7bit
Content-Description: 

Re: Need to clone machines efficiently - help?

[Mircea Luca]

Erik Winn wrote:
> 
> Hi Folks,
> 
>  I have just started working with a group here in Portland that is taking in
> old machines and recycling them - putting linux on as the OS (of course ;}).
> See http://www.freegeek.org for more. Its a non-profit all volunteer thing;
> and actually one of the people has posted to one of these lists before ...
> but, apparently things kinda (lamely, IMO) drifted toward a mandrake system
> -- I think I can turn that around with a little help; I am putting in some
> good time there and I think that when the realities of maintaining and
> upgrading rpm systems hits they may change their minds.
> 
>  Here is the first obstacle - not really a big one, but I spent all day
> digging around and couldn't really find any tools for this one: we want to be
> able to clone the machines easily over the local net. Mandrake has a tricky
> boot floppy that asks only for the eth0 config and then runs a bunch of perl
> to do the rest of the install non-interactively. I haven't started reading
> the scripts yet (that's plan B), instead I was hoping that someone had come
> up with something similar for debian. We are looking at hundreds of boxes
> already and its really just begun.
> 
>  This is really not a huge task - it would just make a nice splash over here
> if we could come up with something ...
> 
>  I would greatly appreciate recieving help/ideas/advice on this - Note
> however that I am not actually subscribed to the list at present (sorry, just
> too much at the moment ) so you can reply to me personally if you like.
> 
>  Thanks very much in advance! I hope we can get this to happen.
> 
> Erik Winn
> 
> --
This is what you want I guess.Made for Debian

http://www.informatik.uni-koeln.de/fai/

-- 
The best way to escape from a problem is to solve it. 
 Alan Saporta

[no subject]

[Red`XIII]

Re: partial mirroring script (that actually doesn't work)

[Joey Hess]

Hi, fwiw the current and generally working version of my partial mirror
script is at http://cvs.kitenet.net/joey-cvs/bin/debmirror

-- 
see shy jo

Re: Location of -doc documentation?

[Joey Hess]

Karl M. Hegbloom wrote:
>  I just noticed that `apache-doc' puts the documentation under
>  "http://.../doc/apache";, while `debconf-doc' puts it under
>  "http://.../debconf-doc/";. 

Eh? (Debconf-doc is a package, that contains some documentation files.
It doesn't touch the web space at all.)

-- 
see shy jo

Re: Location of -doc documentation?

[Arthur Korn]

Joey Hess schrieb:
> Karl M. Hegbloom wrote:
> >  I just noticed that `apache-doc' puts the documentation under
> >  "http://.../doc/apache";, while `debconf-doc' puts it under
> >  "http://.../debconf-doc/";. 
> 
> Eh? (Debconf-doc is a package, that contains some documentation files.
> It doesn't touch the web space at all.)

Well, it touches /usr/doc, and /etc/apache/srm.conf has an Alias
/doc/ /usr/doc/. HTH

ciao, 2ri

Re: What do you wish for in an package manager?

[Arthur Korn]

Hi

Mark Seaborn schrieb:
> I want a system where I can install multiple versions of a library (or
> any package really) and say which version I want each program on the
> system to use, possibly on a per-user basis.  The present system is a
> disaster waiting to happen:  If I install a package from unstable, it
> often wants to replace my version of libc from stable with one from
> unstable.
[...]

You actually can install (hypotetical) libfoo0
(/usr/lib/libfoo.so.0.3.1) and libfoo1
(/usr/lib/libfoo.so.1.0.9) at once, and that's why Debians
shared library dependencies work (with packages gradually
upgrading to the new library). Unfortunately more and more
library packages do no more properly feature the entire soname
in the package name, which can cause mayham.

And if you want to install packages from unstable on a stable
system you ether take the binary package and everything it
depends on, or you apt-get -b source it. If all library packages
are made properly, you can't get around this with fancy package
management.

Running programs with another than the standard version of a
library can be done with LD_PRELOAD (ld.so(8)).

ciao, 2ri

Re: Location of -doc documentation?

[Nathan E Norman]

On Mon, Dec 25, 2000 at 11:48:32PM +0100, Arthur Korn wrote:
> Joey Hess schrieb:
> > Karl M. Hegbloom wrote:
> > >  I just noticed that `apache-doc' puts the documentation under
> > >  "http://.../doc/apache";, while `debconf-doc' puts it under
> > >  "http://.../debconf-doc/";. 
> > 
> > Eh? (Debconf-doc is a package, that contains some documentation files.
> > It doesn't touch the web space at all.)
> 
> Well, it touches /usr/doc, and /etc/apache/srm.conf has an Alias
> /doc/ /usr/doc/. HTH

I beleive the original point was that debconf-doc places its
documentation in /usr/share/doc/debconf-doc, while apache-doc places
its documentation in /usr/share/doc/apache, rather than
/usr/share/doc/apache-doc.  (Principle of least surprise, I suppose).

Cheers,

-- 
Nathan Norman - Staff Engineer | A good plan today is better
Micromuse Inc. | than a perfect plan tomorrow.
mailto:[EMAIL PROTECTED]   |   -- Patton


pgp1nRNDcdfpa.pgp
Description: PGP signature

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Brian May]

> "Russell" == Russell Coker <[EMAIL PROTECTED]> writes:

Russell> On Saturday 23 December 2000 09:13, KORN Andras wrote:
>> I feel that there exists a general confusion among some Debian
>> developers as to what user ids such as 'nobody' should be used
>> for. I suggest that the policy be updated with relevant advice.

Russell> Nobody should never be used.  If you use nobody then
Russell> someone else will choose to use it for the same reasons
Russell> and you end up with two programs sharing the same UID.
Russell> The only solution is to have nothing use it as a matter
Russell> of policy.

Thats my opinion too. Any process run as "nobody" can be controlled by
another process run by "nobody" that has been compromised, via
signals, looking for secrets in core dump files, strace, gdb,
etc. (strace and gdb can both attach to a running program).

However, the idea of one UID per daemon is (IMHO) a really horrible
solution, too, as you end up having more UIDs for daemons then
users. The best solution, capabilities, is yet to be implemented in
the relevant software.


As for the issue that www-data shouldn't own any data files (now that
is a contradiction in names), that is less clear cut. People want
web pages to be

a) private, so access can be controlled via apache.
b) editable by anyone in the www-data group can make changes.
c) read-only to the web server.

which is a conflicting list of goals unless ACLs are supported.
-- 
Brian May <[EMAIL PROTECTED]>

Re: What do you wish for in an package manager?

[Brian May]

> "Dwayne" == Dwayne C Litzenberger <[EMAIL PROTECTED]> writes:

Dwayne> Hello!  I'm starting work on a new linux package manager.
Dwayne> The idea is to be able to replace rpm, dpkg, apt, dselect
Dwayne> (backend) with one,written mostly from scratch and
Dwayne> designed to be as simple (code, not features) and clean as
Dwayne> possible.  For now, the work will be strictly academic,
Dwayne> but if it works out, it may evolve into future standard
Dwayne> package manager.

Dwayne> So my question is: What do you wish for in a package
Dwayne> manager?

1. Built in support for shared NFS systems.

See http://snoopy.apana.org.au/~bam/debian/nfs-dpkg/> for
some examples.


2. Get rid of maintainer scripts (don't ask me how...) so that
upgrading packages is guaranteed not to destroy your computer, even if
the package came an from untrusted source. This could be carried
further by saying "no daemons can be started by UID=root without
express permission by some protected config file". Perhaps maintainer
scripts can run from a chroot and/or non-root environment (issues
remain unsolved).


3. Other minor issues, eg prevent a package from getting purged but
accidently leaving files behind for whatever reason, etc.


Actually, but doing 2 you will automatically solve one of the biggest
issues in solving 1.
-- 
Brian May <[EMAIL PROTECTED]>

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Hamish Moffatt]

On Tue, Dec 26, 2000 at 11:13:13AM +1100, Brian May wrote:
> However, the idea of one UID per daemon is (IMHO) a really horrible
> solution, too, as you end up having more UIDs for daemons then
> users. 

Why is that a problem? There are 65536 available UIDs.


Hamish
-- 
Hamish Moffatt VK3SB <[EMAIL PROTECTED]> <[EMAIL PROTECTED]>

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Anand Kumria]

On Tue, Dec 26, 2000 at 11:48:35AM +1100, Hamish Moffatt wrote:
> On Tue, Dec 26, 2000 at 11:13:13AM +1100, Brian May wrote:
> > However, the idea of one UID per daemon is (IMHO) a really horrible
> > solution, too, as you end up having more UIDs for daemons then
> > users. 
> 
> Why is that a problem? There are 65536 available UIDs.
> 

And 2^32 when 32-bit UIDs become more widespread.

Anand

-- 
Linux.Conf.Au   --  http://linux.conf.au/
17th - 20th January,--  Alan Cox, David Miller,
Sydney, Australia   --  Tridge, maddog and you?

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Brian May]

> "Hamish" == Hamish Moffatt <[EMAIL PROTECTED]> writes:

Hamish> On Tue, Dec 26, 2000 at 11:13:13AM +1100, Brian May wrote:
>> However, the idea of one UID per daemon is (IMHO) a really
>> horrible solution, too, as you end up having more UIDs for
>> daemons then users.

Hamish> Why is that a problem? There are 65536 available UIDs.

Well yes and no. On most desktop systems there never will be a problem.

Some potential problems though:

- easy to hide back-door entry point in /etc/passwd if lots of entries
exist (eg. missing password field). Whether this is by mistake
or done on purpose by an attacker is not important, but the fact it
is harder to detect may be important.

- As the number of entries grows, the chance that one/more entries
will conflict with some NIS, openldap or remote NFS system increases.
Especially since adduser, etc, do not support NIS or openldap.  I am
not sure of the details here - can adduser assign a local user a UID
that conflicts with that from some other source?

- harder to administrate /etc/passwd as more users exist.
-- 
Brian May <[EMAIL PROTECTED]>

Re: Misclassification of packages; "libs" and "doc" sections

[Eray Ozkural \(exa\)]

Hi Thomas,

I got back to working on ontology, and I'd like to give an answer to
one of your previous remarks. Your last e-mail was a bit harsh but
I'm hoping that you will find my view worthwhile. ;)

"Thomas Bushnell, BSG" wrote:
> 
> I think that logic has a great deal to do with semantics.
> 
> I think that the mathematical notion Category Theory has a great deal
> to do with logic and semantics.
> 
> And I don't think that the mathematical notion "Category Theory" has a
> great deal to do with the traditional philosophical notion of
> "Category".

As I read the views of Category Theorists themselves, I found out
that their views were quite parallel with yours. They thought of
metaphysics to be a distinct sphere of endeavor from that of mathematics
as is the general attitude of a mathematician. In countless encounters,
mathematicians did admit that they held their task as a subjective and
intuitive effort rather than an objective "hard" science and that
it had a boundary of its own. That its use did not necessarily indicate
a relation with another field. I may use mathematics to explain a
sociological phenomenon, but mathematics is not related to sociology;
it remains separate. I believe most would claim such even for physics
which itself has given way to new branches of mathematics.

Anyway, mathematicians would mostly tell me
  mathematics != philosophy
in strong words.

Unfortunately, I have to disagree :)

The creators of mathematical theories, which have been used in AI
to explain how facts of world are to be represented, would say
that the terms they have used are only borrowed. *The semantic content
is entirely different* That I believe is also what you have
stated in various ways. That the term "category" in Category Theory
is very different from Category in study of ontology in philosophy.

That is precisely where I think we *should* be skeptical about. After
thinking about Feyerabend's view of scientific practice, I reached
the following argument:

  Artificially separating mutually related theories into predefined
domains of theories is not a fruitful methodology.

It isn't because it is the enforcement of a particular "progress"
methodology. That it is the correct way to draw a hard line between
the philosophy and science of "a thing". I cannot offer a proof of
my argument why this methodology is wrong, because it has to be a
"historical" rather than an analytical one. That I am not very skilled at;
and I have limited space in this e-mail.

In the context of our discussion, I think my argument would imply
something of the following sort:
  We should not exclude the possibility that a "new theory" of
categories, be it more metaphysical or more mathematical, depends
on an important relation between the mathematical Category and
philosophical Category regardless of whether prominent philosophers
or mathematicians deny such a relation.

That was a cumbersome sentence, but I can't word it better now :)

The simpler statement would be that in new research we should
utilize ideas from both worlds and try to exploit similarities
as well as differences between them. I know this sounds confusing
but I think it is an "anything goes" argument for research. We should
not inhibit scientific practice before it happens.

I've written this small piece because I was inspired by a work
of Nicola Guarino. If you're interested I can send you links
to some of his papers. Of course you can take a look at his work
yourself. At this location there are many papers authored by him:

http://www.ladseb.pd.cnr.it/infor/ontology/Papers/OntologyPapers.html

Merry Christmas,

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: What do you wish for in an package manager?

[Eray Ozkural \(exa\)]

Brian May wrote:
> 
> 2. Get rid of maintainer scripts (don't ask me how...) so that
> upgrading packages is guaranteed not to destroy your computer, even if
> the package came an from untrusted source. This could be carried
> further by saying "no daemons can be started by UID=root without
> express permission by some protected config file". Perhaps maintainer
> scripts can run from a chroot and/or non-root environment (issues
> remain unsolved).

Won't ask you how :) Here's a MFTL sol'n :)

You need to devise a package description/configuration language
that is declarative rather than procedural.

What comes to my mind would be some sort of "logical language", maybe
something based on Prolog. That the statements as your example would
be implemented with it and then the package interpreter would
handle the "procedural" aspects of upgrading.

No religious wars, all right? :)

Cheers,

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Oops

[Eray Ozkural \(exa\)]

Mistakenly sent to debian-devel. This is off topic.

Merry Xmas to you all!!

Cheers,

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Eray Ozkural \(exa\)]

Brian May wrote:
> 
> - harder to administrate /etc/passwd as more users exist.

I like using groups to give different sets of rights and I'm
annoyed by Debian giving every user his own group. Is that
reall necessary?

cu,

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: Misclassification of packages; "libs" and "doc" sections

[Anand Kumria]

On Tue, Dec 26, 2000 at 04:29:22AM +0200, Eray Ozkural (exa) wrote:
> Hi Thomas,
> 

[snip - lengthy email]

All of which is fine and dandy. None of which belonged on a public
mailing list though. Your email is tangentially related to Debians'
effort to determine new categories. 

In future please send those kinds of emails privately.

Anand

-- 
Linux.Conf.Au   --  http://linux.conf.au/
17th - 20th January,--  Alan Cox, David Miller,
Sydney, Australia   --  Tridge, maddog and you?

Re: Misclassification of packages; "libs" and "doc" sections

[Eray Ozkural \(exa\)]

Anand Kumria wrote:
> 
> In future please send those kinds of emails privately.

mis-take. :)

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: What do you wish for in an package manager

[Jeffry Smith]

Another thing I would like is something like the BSD "ports" -
download the source, have my machine do the compile, but still have
all the dependencies properly worked out (sort of an expanded apt-get
-b source).


-- 
jeff smith
-
thought for the day:  He that composes himself is wiser than he that
composes a book.
-- B. Franklin

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Brian May]

> "exa" == exa   writes:

exa> Brian May wrote:
>> - harder to administrate /etc/passwd as more users exist.

exa> I like using groups to give different sets of rights and I'm
exa> annoyed by Debian giving every user his own group. Is that
exa> reall necessary?

I don't do that on my machine here. Just edit /etc/adduser.conf
Previously you had to be careful that the default umask was setup
correctly, not sure if this is an issue or not now.
-- 
Brian May <[EMAIL PROTECTED]>

Re: What do you wish for in an package manager?

[Brian May]

> "exa" == exa   writes:

exa> You need to devise a package description/configuration
exa> language that is declarative rather than procedural.

exa> What comes to my mind would be some sort of "logical
exa> language", maybe something based on Prolog. That the
exa> statements as your example would be implemented with it and
exa> then the package interpreter would handle the "procedural"
exa> aspects of upgrading.

Sounds like a good idea to me.

Just come up with a language that:

a) is not dangerous in anyway.
b) has the minimum features required for all packages.

This might not be possible for all packages (consider ssh which
automatically generates a key pair on initial installation). Perhaps a
chroot and/or non-root environment could be setup to handle these
cases?
-- 
Brian May <[EMAIL PROTECTED]>

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Nathan E Norman]

On Tue, Dec 26, 2000 at 04:43:53AM +0200, Eray Ozkural (exa) wrote:
> Brian May wrote:
> > 
> > - harder to administrate /etc/passwd as more users exist.
> 
> I like using groups to give different sets of rights and I'm
> annoyed by Debian giving every user his own group. Is that
> reall necessary?

It's useful when you're in a development environment where you've got
directories that are group writable.

On the other hand, I'd guess most large-scale development projects
now use CVS rather than group writable directories as a sharing
mechanism.

FWIW when I was a sysadmin I generally put all untrusted users in a
single group (or divided them into classes of groups).

Regards,

-- 
Nathan Norman - Staff Engineer | A good plan today is better
Micromuse Inc. | than a perfect plan tomorrow.
mailto:[EMAIL PROTECTED]   |   -- Patton


pgpjCOCKfltIo.pgp
Description: PGP signature

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Eray Ozkural \(exa\)]

Nathan E Norman wrote:
> 
> On Tue, Dec 26, 2000 at 04:43:53AM +0200, Eray Ozkural (exa) wrote:
> > I like using groups to give different sets of rights and I'm
> > annoyed by Debian giving every user his own group. Is that
> > reall necessary?
> 
> It's useful when you're in a development environment where you've got
> directories that are group writable.
> 
> On the other hand, I'd guess most large-scale development projects
> now use CVS rather than group writable directories as a sharing
> mechanism.

I put CVS users in a group called developers. Is that wrong?

Thanks,

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Eray Ozkural \(exa\)]

Brian May wrote:
> 
> > "exa" == exa   writes:
> 
> exa> Brian May wrote:
> >> - harder to administrate /etc/passwd as more users exist.
> 
> exa> I like using groups to give different sets of rights and I'm
> exa> annoyed by Debian giving every user his own group. Is that
> exa> reall necessary?
> 
> I don't do that on my machine here. Just edit /etc/adduser.conf
> Previously you had to be careful that the default umask was setup
> correctly, not sure if this is an issue or not now.

Yep. I discovered that umask issue. I guess it's still a problem.

Thanks,

-- 
Eray (exa) Ozkural
Comp. Sci. Dept., Bilkent University, Ankara
e-mail: [EMAIL PROTECTED]
www: http://www.cs.bilkent.edu.tr/~erayo

Re: Need to clone machines efficiently - help?

[Peter Eckersley]

On Mon, Dec 25, 2000 at 12:15:50AM -0800, Erik Winn wrote:

> Hi Folks,
> 
>  I have just started working with a group here in Portland that is taking in 
> old machines and recycling them - putting linux on as the OS (of course ;}). 
> See http://www.freegeek.org for more. Its a non-profit all volunteer thing; 
> and actually one of the people has posted to one of these lists before ... 
> but, apparently things kinda (lamely, IMO) drifted toward a mandrake system 
> -- I think I can turn that around with a little help; I am putting in some 
> good time there and I think that when the realities of maintaining and 
> upgrading rpm systems hits they may change their minds.

Cool...

we've got an established organisation doing this in Australia now:

http://www.computerbank.org.au

We're using Debian for all of our systems.

> 
>  Here is the first obstacle - not really a big one, but I spent all day 
> digging around and couldn't really find any tools for this one: we want to be 
> able to clone the machines easily over the local net. Mandrake has a tricky 
> boot floppy that asks only for the eth0 config and then runs a bunch of perl 
> to do the rest of the install non-interactively. I haven't started reading 
> the scripts yet (that's plan B), instead I was hoping that someone had come 
> up with something similar for debian. We are looking at hundreds of boxes 
> already and its really just begun.

This was extremely difficult for us at first, simply because the
hardware we got was so variable that no "standard install" was really
possible.  There was one small shipment of machines with identical disks
which we cloned using dd :).

Things have got much better lately, since we started receiving corporate
donations of largeish groups of modern PCs with similar hardware.  The
way we've ended up doing it is this:

* A debian mirror server
* Customised task packages

So we start a normal debian install, but then pick
task-computerbank-whatever and it's done.

A custom task package might play really well with an automated
installer, if your hardware is sufficiently uniform to support one.

> 
>  This is really not a huge task - it would just make a nice splash over here 
> if we could come up with something ...
> 
>  I would greatly appreciate recieving help/ideas/advice on this - Note 
> however that I am not actually subscribed to the list at present (sorry, just 
> too much at the moment ) so you can reply to me personally if you like.
> 
>  Thanks very much in advance! I hope we can get this to happen.
> 

Good luck with your project.  Take a look at Computerbank, and if the
goals are close enough, perhaps we should start considering
international affiliations (Computerbank has several chapters in
different cities, and we've found that this has certainly been to our
advantage - international links would be even better).

-- 

|> |= -+- |= |>
|  |-  |  |- |\

Peter Eckersley
([EMAIL PROTECTED])
http://www.cs.mu.oz.au/~pde

for techno-leftie inspiration, take a look at
http://www.computerbank.org.au/


pgp4lKXf5Bgzz.pgp
Description: PGP signature

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Brian May]

> "Eray" == Eray Ozkural exa <[EMAIL PROTECTED]> writes:

Eray> Yep. I discovered that umask issue. I guess it's still a
Eray> problem.

zsh has in /etc/zshrc:

[[ $UID == $GID ]] && umask 002 || umask 022

My only dislike is it overrides my default setup in ~/.zshenv of 077.
It seems wrong to put this stuff in zshrc, that only gets used for
interactive shells. zshenv gets processed for all shells, but is run
before zshrc.
-- 
Brian May <[EMAIL PROTECTED]>

Re: Bug#80343: general: Lack of policy on which files should be owned by which user

[Hamish Moffatt]

On Tue, Dec 26, 2000 at 04:43:53AM +0200, Eray Ozkural (exa) wrote:
> I like using groups to give different sets of rights and I'm
> annoyed by Debian giving every user his own group. Is that
> reall necessary?

No, but it's a good idea. It makes it much easier to work in
directories shared with other users (but not all users), because
you don't have to keep changing your umask all the time, or
even worse, fixing file permissions because you (or somebody
else) forgot to change their umask.

What's the harm in it?


Hamish
-- 
Hamish Moffatt VK3SB <[EMAIL PROTECTED]> <[EMAIL PROTECTED]>