RSS feed

Re: How to configure for delegated authentication

[Date Prev][Date Next] [Thread Prev][Thread Next]

Re: How to configure for delegated authentication

On Tue, Oct 04, 2011 at 05:22:48PM -0400, Jeff Mitchell wrote:
> Hello,
> I have a service (git, via gitolite) that I'd like to make available to
> users. The problem is that we have about 4k users and I'd like not to
> have to create and synchronize local accounts on the box for each one.
> I'm currently having two issues:
> 1) The best way for me to do this would be to perform what's often
> called "delegated authentication" against LDAP. Instead of going to LDAP
> to look up the user credentials, I need to test the given user
> credentials by doing a bind against LDAP with those given credentials

As near as I can tell, nslcd already tests user login in by binding as that 
user (at least the way it's set up in Debian). I don't have binddn/bindpw 
configured in /etc/nslcd.conf, and my OpenLDAP ACLs disallow "read" on the 
userPassword attribute.

> (or even better, with a given binddn/bindpw but then do a rebind, I
> think it's called, with the given user credentials -- it's what many web
> apps do). Our LDAP server itself goes to a third party to validate
> credentials, so the LDAP server does not have password information,
> hence why I need to do authentication based upon bind results. I'm not
> sure how I can do this, or if I can do this...I'm happy to try to help
> implement it if needed, but I'd probably need some hand-holding.

This sounds like the translucent proxy in OpenLDAP:

> 2) I would need to return a particular shell for the user in order to
> continue with the git functionality. From the earlier nss_ldap I should
> have been able to use the nss_override_attribute_value command, but from
> the current sources it looks like that's been removed...?

If these aren't shell users, why not change the loginShell for the entry to be 
whatever will enable git? Either in the proxy config or in the remote ldap 
server. (Your mileage may vary.)

> Thanks,
> Jeff
> -- 
> To unsubscribe send an email to
> or see
To unsubscribe send an email to or see