svnserve passwd plaintext

[gmu 2k6 <gm...@gmail.com>]

is there no way to use crypt(), md5(), hashXY(), bcrypt() for password
enrires in the passwd file when using svnserve? using mod_dav_svn or
tunneling svn via SSH is no option right now.
if there really is no way I might fall back to writing a passwd
management tool which works with base64 passwords but of course writes
them debase64'ed into passwd. this way at I would have to actievely
take a peek at passwd to see the passwords.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: svnserve passwd plaintext

[gmu 2k6 <gm...@gmail.com>]

On 7/15/06, David Anderson <da...@calixo.net> wrote:
> * gmu 2k6 <gm...@gmail.com> [2006-07-15 12:06:57]:
> > is there no way to use crypt(), md5(), hashXY(), bcrypt() for password
> > enrires in the passwd file when using svnserve? using mod_dav_svn or
> > tunneling svn via SSH is no option right now.
> > if there really is no way I might fall back to writing a passwd
> > management tool which works with base64 passwords but of course writes
> > them debase64'ed into passwd. this way at I would have to actievely
> > take a peek at passwd to see the passwords.
>
> This issue is tackled in the FAQ: the authentication method (cram-md5)
> permits having secure, tamper-proof authentication over an insecure
> channel, at the expense of requiring storage of plaintext passwords on
> the server side.
>
> It has been our long time policy to not offer any obfuscation method,
> as security through obscurity is just an artificial way of falsely
> raising your trust in part of the system, just as it was in CVS' day.

as long as I have to administer a passwd file with plaintext entries I
will need a tool anyway, which allows the users to change their
passwords and me to not touch the file with an editor. this way I
won't need to see the passwords as base64 can be used as a method to
obfuscate it for the common human brain so far that one won't be able
to decode it instantly (individuals being able to tell you the sqrt of
123876876123 put aside of course :D).

> However, you will note that the more complete solution to this problem
> - supporting svn:// over SSL and implementing full support for a SASL
> library - is being worked on right now.

Nice. Plus the SASL work will make svnserve that much more useful:
http://svn.haxx.se/dev/archive-2006-07/0300.shtml

> With this solution, you will have many authentication methods
> available, most of which do not store passwords as plaintext on the
> server (but that do require the secure communications layer provided
> by SSL, as the authentication protocols themselves are susceptible to
> attack).

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: svnserve passwd plaintext

[David Anderson <da...@calixo.net>]

* gmu 2k6 <gm...@gmail.com> [2006-07-15 12:06:57]:
> is there no way to use crypt(), md5(), hashXY(), bcrypt() for password
> enrires in the passwd file when using svnserve? using mod_dav_svn or
> tunneling svn via SSH is no option right now.
> if there really is no way I might fall back to writing a passwd
> management tool which works with base64 passwords but of course writes
> them debase64'ed into passwd. this way at I would have to actievely
> take a peek at passwd to see the passwords.

This issue is tackled in the FAQ: the authentication method (cram-md5)
permits having secure, tamper-proof authentication over an insecure
channel, at the expense of requiring storage of plaintext passwords on
the server side.

It has been our long time policy to not offer any obfuscation method,
as security through obscurity is just an artificial way of falsely
raising your trust in part of the system, just as it was in CVS' day.

However, you will note that the more complete solution to this problem
- supporting svn:// over SSL and implementing full support for a SASL
library - is being worked on right now.

With this solution, you will have many authentication methods
available, most of which do not store passwords as plaintext on the
server (but that do require the secure communications layer provided
by SSL, as the authentication protocols themselves are susceptible to
attack).

- Dave

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: svnserve passwd plaintext

[Erik Huelsmann <eh...@gmail.com>]

On 7/15/06, Nico Kadel-Garcia <nk...@comcast.net> wrote:
> gmu 2k6 wrote:
> > is there no way to use crypt(), md5(), hashXY(), bcrypt() for password
> > enrires in the passwd file when using svnserve? using mod_dav_svn or
> > tunneling svn via SSH is no option right now.
> > if there really is no way I might fall back to writing a passwd
> > management tool which works with base64 passwords but of course writes
> > them debase64'ed into passwd. this way at I would have to actievely
> > take a peek at passwd to see the passwords.
>
> I've been growsing about that: some folks made it clear that the
> communication over svnserver to handle passwords is actually encrypted, but
> I loathe having plain-text passwords lyng around.

Base64 encoding isn't encrypting: it's merely hiding (as is CVS's
ROT13 method)...

bye,

Erik.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org

Re: svnserve passwd plaintext

[Nico Kadel-Garcia <nk...@comcast.net>]

gmu 2k6 wrote:
> is there no way to use crypt(), md5(), hashXY(), bcrypt() for password
> enrires in the passwd file when using svnserve? using mod_dav_svn or
> tunneling svn via SSH is no option right now.
> if there really is no way I might fall back to writing a passwd
> management tool which works with base64 passwords but of course writes
> them debase64'ed into passwd. this way at I would have to actievely
> take a peek at passwd to see the passwords.

I've been growsing about that: some folks made it clear that the 
communication over svnserver to handle passwords is actually encrypted, but 
I loathe having plain-text passwords lyng around.

Instead, I use HTTPS/mod_dav_svn access, which can use .htaccess and other 
user authentication techniques, and svnperms.conf and svnperms.py to control 
user-based read/write/update permissions. Then you can use webmin, htpasswd, 
or other similar tools to edit individual .htaccess accounts. 

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org