Re: Looking for a co-maintainer for adduser

2003-10-08 Thread Joey Hess
John Belmonte wrote:
 All those points are well taken.  I was trying to pull a fast one: 
 surely no one would notice a few files added to base which total 1/3rd 
 the size of bash itself.  But I've been caught.

I hope you realize that the perl-base package is itself hardly larger
than bash.

-- 
see shy jo




Re: Looking for a co-maintainer for adduser

2003-10-06 Thread Fabien Ninoles
Only on a well-written OS... ;)
Mark Brown wrote:
On Mon, Oct 06, 2003 at 01:01:52AM +0200, Marc Haber wrote:

Well-written C++ using well tested class libraries tend to do a pretty
good job, security-wise.

I often find that well written code does a good job.



Re: Looking for a co-maintainer for adduser

2003-10-06 Thread Gunnar Wolf
John Belmonte dijo [Sun, Oct 05, 2003 at 10:20:59AM -0400]:
 Lua is a modern high-level language.  Its 15K stand-alone interpreter 
 depends on only two libraries which total less than 200K.  The 
 functionality of its standard libraries are limited by ANSI C, but there 
 are are third party libraries for talking to the OS, such as a POSIX 
 library.
 
 Someone less averse to prefix notation than I might make the same 
 argument about scsh.  It's larger and has more dependencies, but on the 
 other hand has full library support for system programming.
 
 Why not consider tiny languages?

Because of how powerful is Perl? Because of the amount of things that
depend on Perl that currently exist and would be a waste of time to
rewrite? Because Perl might be the best tool for many cases? There are
many possible answers...

Not that coding in Lua, scsh or similar tools is bad. Not that the
regular shells are not up to the task in many cases. Simply that... Perl
seems to be such a powerful language that people would end up installing
it anyway - And if you have an optional package installed in 95% of the
machines, then you have something I would like calling 'base'.

Greetings,

-- 
Gunnar Wolf - [EMAIL PROTECTED] - (+52-55)5630-9700 ext. 1366
PGP key 1024D/8BB527AF 2001-10-23
Fingerprint: 0C79 D2D1 2C4E 9CE4 5973  F800 D80E F35A 8BB5 27AF




Re: Looking for a co-maintainer for adduser

2003-10-06 Thread John Belmonte
Gunnar Wolf wrote:
Because of how powerful is Perl? Because of the amount of things that
depend on Perl that currently exist and would be a waste of time to
rewrite? Because Perl might be the best tool for many cases? There are
many possible answers...
Not that coding in Lua, scsh or similar tools is bad. Not that the
regular shells are not up to the task in many cases. Simply that... Perl
seems to be such a powerful language that people would end up installing
it anyway - And if you have an optional package installed in 95% of the
machines, then you have something I would like calling 'base'.
Excuse me, but I'm not arguing Lua vs. Perl, or tiny languages vs. Perl. 
 Rather I'm challenging Colin Watson's statement, Take away Perl and 
you've got only shell, C, and C++ left.


--
http:// if   le.o  /



Re: Looking for a co-maintainer for adduser

2003-10-06 Thread Joel Baker
On Mon, Oct 06, 2003 at 03:14:24PM -0500, Gunnar Wolf wrote:
 John Belmonte dijo [Sun, Oct 05, 2003 at 10:20:59AM -0400]:
  Lua is a modern high-level language.  Its 15K stand-alone interpreter 
  depends on only two libraries which total less than 200K.  The 
  functionality of its standard libraries are limited by ANSI C, but there 
  are are third party libraries for talking to the OS, such as a POSIX 
  library.
  
  Someone less averse to prefix notation than I might make the same 
  argument about scsh.  It's larger and has more dependencies, but on the 
  other hand has full library support for system programming.
  
  Why not consider tiny languages?
 
 Because of how powerful is Perl? Because of the amount of things that
 depend on Perl that currently exist and would be a waste of time to
 rewrite? Because Perl might be the best tool for many cases? There are
 many possible answers...

Actually, TTBOMK, it's more or less a historic choice. Which is to say,
at the time the decision was made, Perl was one of the only tools which
COULD do all of the things needed for setting up a base system, in a single
language which could (reasonably) easily express this (bash shell does not
easily express it, even if it *can* express it, generally).

If the decision were being made today, we might see arguments over Perl,
Lua, Python, and probably others - valid ones, in fact. However, now isn't
then, and once the decision has been made, it becomes a much harder thing
to move away from it (for good reason) - since it involves rewriting
everything, with the attendant bugs, problems, unrest, and insanity that
comes with.

(I like Perl; it's my slap-it-together language of choice. Conversely, I
also regularly work on a system where core pieces are written in python,
because that's the religion of the person who set it up. If you want to see
my actual opinion, go read the 'Perl vs Python' bit in Unix Power Tools,
3rd edition - written by the same person, and reviewed in-house to ensure a
fair hearing to Perl :)

Actually, I'd almost prefer Python for doing Debian base work - not because
Perl is bad at it, per se, but because so many people have bad Perl habits,
and this propagates into the scripts. Python makes it (slightly) harder
to write truly gross code, and since Debian maintainer scripts really
shouldn't be doing the stupid tricks that prevents you from doing *anyway*,
I'd be willing to take that loss. But it just isn't worth it to try to
convert everything (at least, not until Perl 6 forces us to consider
rewriting it all, anyway...)

Though getting Python to have an easier Build-Depend chain would be nice
for porters, if we ever do support it as an option for maintainer scripts.
-- 
Joel Baker [EMAIL PROTECTED],''`.
Debian GNU NetBSD/i386 porter: :' :
 `. `'
   `-


pgppklYb4P64n.pgp
Description: PGP signature


Re: Looking for a co-maintainer for adduser

2003-10-06 Thread Colin Watson
On Mon, Oct 06, 2003 at 05:42:05PM -0400, John Belmonte wrote:
 Gunnar Wolf wrote:
 Because of how powerful is Perl? Because of the amount of things that
 depend on Perl that currently exist and would be a waste of time to
 rewrite? Because Perl might be the best tool for many cases? There are
 many possible answers...
 
 Not that coding in Lua, scsh or similar tools is bad. Not that the
 regular shells are not up to the task in many cases. Simply that... Perl
 seems to be such a powerful language that people would end up installing
 it anyway - And if you have an optional package installed in 95% of the
 machines, then you have something I would like calling 'base'.
 
 Excuse me, but I'm not arguing Lua vs. Perl, or tiny languages vs. Perl. 
  Rather I'm challenging Colin Watson's statement, Take away Perl and 
 you've got only shell, C, and C++ left.

You haven't challenged it successfully, then; to my knowledge, my
statement is correct for the current base system, which is what it was
referring to.

I don't necessarily oppose tiny languages such as Lua, but perhaps
somebody should write the tools in question in them first, otherwise
this is pure vapourware. (Or, then again, perhaps they shouldn't; I'm
not a fan of rewriting things that already work.)

Cheers,

-- 
Colin Watson  [EMAIL PROTECTED]




Re: Looking for a co-maintainer for adduser

2003-10-06 Thread John Belmonte
Colin Watson wrote:
You haven't challenged it successfully, then; to my knowledge, my
statement is correct for the current base system, which is what it was
referring to.
I don't necessarily oppose tiny languages such as Lua, but perhaps
somebody should write the tools in question in them first, otherwise
this is pure vapourware. (Or, then again, perhaps they shouldn't; I'm
not a fan of rewriting things that already work.)
All those points are well taken.  I was trying to pull a fast one: 
surely no one would notice a few files added to base which total 1/3rd 
the size of bash itself.  But I've been caught.

As for the evils of rewriting, I tend to agree.  However, in this case, 
the maintainer of adduser said, ... I don't like the way adduser is 
currently written (and also perl) a lot, and was planning to do a
complete overhaul, which prompted me to be so bold.


--
http:// if   le.o  /



Re: Looking for a co-maintainer for adduser

2003-10-05 Thread Marc Haber
On Sat, 04 Oct 2003 15:58:46 -0500, John Hasler [EMAIL PROTECTED]
wrote:
But it would not be nice to not have the things that would leave with it.

Generally, it would be a good thing to have Debian base installable
without perl. That way, security-aware administrators would have the
right to choose whether they need $BELL or $WHISTLE that makes perl
appear.

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: Looking for a co-maintainer for adduser

2003-10-05 Thread Colin Watson
On Sun, Oct 05, 2003 at 01:41:56PM +0200, Marc Haber wrote:
 On Sat, 04 Oct 2003 15:58:46 -0500, John Hasler [EMAIL PROTECTED]
 wrote:
 But it would not be nice to not have the things that would leave with it.
 
 Generally, it would be a good thing to have Debian base installable
 without perl. That way, security-aware administrators would have the
 right to choose whether they need $BELL or $WHISTLE that makes perl
 appear.

I'd rather that the tools in Debian base were written in a high-level
language where available. Take away Perl and you've got only shell, C,
and C++ left; I don't think that's going to improve security in
practice.

-- 
Colin Watson  [EMAIL PROTECTED]




Re: Looking for a co-maintainer for adduser

2003-10-05 Thread John Belmonte
Colin Watson wrote:
I'd rather that the tools in Debian base were written in a high-level
language where available. Take away Perl and you've got only shell, C,
and C++ left; I don't think that's going to improve security in
practice.
Lua is a modern high-level language.  Its 15K stand-alone interpreter 
depends on only two libraries which total less than 200K.  The 
functionality of its standard libraries are limited by ANSI C, but there 
are are third party libraries for talking to the OS, such as a POSIX 
library.

Someone less averse to prefix notation than I might make the same 
argument about scsh.  It's larger and has more dependencies, but on the 
other hand has full library support for system programming.

Why not consider tiny languages?

--
http:// if   le.o  /



Re: Looking for a co-maintainer for adduser

2003-10-05 Thread Scott James Remnant

On Sun, 2003-10-05 at 12:41, Marc Haber wrote:
 On Sat, 04 Oct 2003 15:58:46 -0500, John Hasler [EMAIL PROTECTED]
 wrote:
 But it would not be nice to not have the things that would leave with it

Re: Looking for a co-maintainer for adduser

2003-10-05 Thread Marc Haber
On Sun, 5 Oct 2003 10:54:50 +0100, Colin Watson [EMAIL PROTECTED]
wrote:
I'd rather that the tools in Debian base were written in a high-level
language where available. Take away Perl and you've got only shell, C,
and C++ left; I don't think that's going to improve security in
practice.

Well-written C++ using well tested class libraries tend to do a pretty
good job, security-wise.

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: Looking for a co-maintainer for adduser

2003-10-05 Thread Mark Brown
On Mon, Oct 06, 2003 at 01:01:52AM +0200, Marc Haber wrote:

 Well-written C++ using well tested class libraries tend to do a pretty
 good job, security-wise.

I often find that well written code does a good job.

-- 
You grabbed my hand and we fell into it, like a daydream - or a fever.




Re: Looking for a co-maintainer for adduser

2003-10-05 Thread Bob Proulx
Marc Haber wrote:
 Colin Watson wrote:
 I'd rather that the tools in Debian base were written in a high-level
 language where available. Take away Perl and you've got only shell, C,
 and C++ left; I don't think that's going to improve security in
 practice.
 
 Well-written C++ using well tested class libraries tend to do a pretty
 good job, security-wise.

Saying well-written is cheating.  Any well written program is always
good by definition or it is not be well written.  But what about
poorly written cruft?  Almost all languages are easy to write badly.
But some are easier than others.  Both C++ and Perl come to my mind
when I think of bad programming practices and swiss army chainsaws.

Bob


pgppvqBwloD23.pgp
Description: PGP signature


Re: Looking for a co-maintainer for adduser

2003-10-05 Thread steve
On Sun, Oct 05, 2003 at 05:18:45PM -0600, Bob Proulx wrote:

 Saying well-written is cheating.  Any well written program is always
 good by definition or it is not be well written.  But what about
 poorly written cruft?  Almost all languages are easy to write badly.
 But some are easier than others.  Both C++ and Perl come to my mind
 when I think of bad programming practices and swiss army chainsaws.

  I think the point is that good code and bad code are possible in any
 language, and the panacea of switching to a particular language and 
 expecting all coding programs to go away is simplistic and unrealistic.

  Sure in some languages like Java there aren't going to be pointer
 problems, but other avenues of attack are just as likely; insecure use
 of temporary files, symlink attacks, signal attacks and etc.

Steve
--
# Debian Security Audit Project
http://www.steve.org.uk/Debian/




Re: Looking for a co-maintainer for adduser

2003-10-04 Thread Marc Haber
On Fri, 03 Oct 2003 18:49:40 +0100, Scott James Remnant
[EMAIL PROTECTED] wrote:
Some of us would like to see Perl taken out of base as well :)

That would be an awfully nice thing to have.

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: Looking for a co-maintainer for adduser

2003-10-04 Thread John Hasler
Scott James Remnant wrote:
 Some of us would like to see Perl taken out of base as well :)

Marc writes:
 That would be an awfully nice thing to have.

But it would not be nice to not have the things that would leave with it.
-- 
John Hasler
[EMAIL PROTECTED] (John Hasler)
Dancing Horse Hill
Elmwood, WI




Re: Looking for a co-maintainer for adduser

2003-10-03 Thread Scott James Remnant
On Thu, 2003-10-02 at 09:49, Colin Watson wrote:

 On Thu, Oct 02, 2003 at 10:16:28AM +0200, Domenico Andreoli wrote:
  i have developed a system in python which abstracts from the backend too.
  moreover it is easily expandable with python plugins. it is also easy to
  develop new applications/utilities using it as a python module. it is
  pretty stable, i already use it in production system.
 
 That would mean we'd have to add python to the base system. Perhaps a
 bit much in size terms? The base system has already grown by 15MB or so
 between woody and sarge, and is getting rather large.
 
Indeed, keep Python out of base.

Some of us would like to see Perl taken out of base as well :)

Scott
-- 
Have you ever, ever felt like this?
Had strange things happen?  Are you going round the twist?


signature.asc
Description: This is a digitally signed message part


Looking for a co-maintainer for adduser

2003-10-02 Thread Roland Bauerschmidt
The number of bugs on the adduser package has constantly increased for
the last few months, though none of them is release critical. Since I
was busy with other stuff (mostly OpenLDAP and related stuff) I didn't
keep up with all the feature requests and non-critical bugs. This is
also partly due to the fact that I don't like the way adduser is
currently written (and also perl) a lot, and was planning to do a
complete overhaul (http://www.hbg-bremen.de/~roland/code/adduser.xhtml).

Matthew Palmer has done some nice work in abstracting the passwd storage
backend, and adding methods for LDAP storage. The latter, though, still
needs some more work to be useful in different directory structures.

I am thus seeking for one or two co-maintainers, and appreciate it a lot
if Matthew would chose to be one of them. The package is managed in a
Subversion repository on Alioth. The main package is in trunk, Matthew's
LDAP extended version in brances/adduser-ldap (which should eventually
be merged into trunk); all in the svn+ssh://svn.debian.org/svn/adduser
repository. It'd be particularly useful, if you have NIS experience (and
maybe even a running setup), but not required. There is an adduser-devel
also on Alioth.

If you are interested, drop me a note off-list.

-- Roland




Re: Looking for a co-maintainer for adduser

2003-10-02 Thread Domenico Andreoli
On Thu, Oct 02, 2003 at 10:02:38AM +0200, Roland Bauerschmidt wrote:
 The number of bugs on the adduser package has constantly increased for
 the last few months, though none of them is release critical. Since I
 was busy with other stuff (mostly OpenLDAP and related stuff) I didn't
 keep up with all the feature requests and non-critical bugs. This is
 also partly due to the fact that I don't like the way adduser is
 currently written (and also perl) a lot, and was planning to do a
 complete overhaul (http://www.hbg-bremen.de/~roland/code/adduser.xhtml).
 
i could be interested

 Matthew Palmer has done some nice work in abstracting the passwd storage
 backend, and adding methods for LDAP storage. The latter, though, still
 needs some more work to be useful in different directory structures.
 
i have developed a system in python which abstracts from the backend too.
moreover it is easily expandable with python plugins. it is also easy to
develop new applications/utilities using it as a python module. it is
pretty stable, i already use it in production system.

http://www.nongnu.org/prua/

 I am thus seeking for one or two co-maintainers, and appreciate it a lot
 if Matthew would chose to be one of them. The package is managed in a
 Subversion repository on Alioth. The main package is in trunk, Matthew's
 LDAP extended version in brances/adduser-ldap (which should eventually
 be merged into trunk); all in the svn+ssh://svn.debian.org/svn/adduser
 repository. It'd be particularly useful, if you have NIS experience (and
 maybe even a running setup), but not required. There is an adduser-devel
 also on Alioth.
 

-[ Domenico Andreoli, aka cavok
 --[ http://filibusta.crema.unimi.it/~cavok/gpgkey.asc
   ---[ 3A0F 2F80 F79C 678A 8936  4FEE 0677 9033 A20E BC50




Re: Looking for a co-maintainer for adduser

2003-10-02 Thread Colin Watson
On Thu, Oct 02, 2003 at 10:16:28AM +0200, Domenico Andreoli wrote:
 On Thu, Oct 02, 2003 at 10:02:38AM +0200, Roland Bauerschmidt wrote:
  Matthew Palmer has done some nice work in abstracting the passwd storage
  backend, and adding methods for LDAP storage. The latter, though, still
  needs some more work to be useful in different directory structures.
 
 i have developed a system in python which abstracts from the backend too.
 moreover it is easily expandable with python plugins. it is also easy to
 develop new applications/utilities using it as a python module. it is
 pretty stable, i already use it in production system.
 
 http://www.nongnu.org/prua/

That would mean we'd have to add python to the base system. Perhaps a
bit much in size terms? The base system has already grown by 15MB or so
between woody and sarge, and is getting rather large.

-- 
Colin Watson  [EMAIL PROTECTED]




Re: Looking for a co-maintainer for adduser

2003-10-02 Thread John Hasler
Colin Watson writes:
 That would mean we'd have to add python to the base system.

I'd _really_ rather not see that.  While I now use Python in preference to
Perl, I don't think its advantages justify bloating base.  Perl's just
another procedural language.
-- 
John Hasler
[EMAIL PROTECTED] (John Hasler)
Dancing Horse Hill
Elmwood, WI