> The current spec is in the case directory as proposal.txt and the 
> updated pam_list(5) man page is in the materials directory. Both are 
> attached here as well.

        I wouldn't say I have an objection to this proposal, just a couple
        questions and nits:

> SYNOPSIS
>      pam_list.so.1

> DESCRIPTION
>      The following options can be passed to the       module:
> 
>      allow=
>       The full pathname to a file of allowed users and/or netgroups.
>         Only one of allow= or deny= may be specified.
> 
>      deny=
>         The full pathaname to a file of denied users and/or netgroups.
>         Only one of deny= or allow= may be specified.
> 
>      user
>         The module should only perform netgroup matches on the username.
> 
>      nouser
>         The username should not be used in the netgroup match.
> 
>      host
>         Only the host should be used in netgroup matches.
> 
>      nohost
>         The hostname should not be used in netgroup matches.
> 
>      user_host_exact
>         The user and hostname must be in the same netgroup.

        I believe an EXAMPLES section with the examples in the
        proposal would help here.

> ATTRIBUTES
>      See attributes(5) for descriptions       of the  following  attri-
>      butes:
> 
>      ____________________________________________________________
>     |     ATTRIBUTE TYPE        |       ATTRIBUTE VALUE       |
>     |_____________________________|_____________________________|
>     | Interface       Stability         | Evolving                    |
                                            ^^^^^^^^
                                            Committed
> 1. Background
> -------------

> If you have complex netgroups and use them with passwd compat mode then
> there can be a noticeable and significant delay on every getpw*() call.
> This is due to the need to walk the netgroup to find the user, this is
> made worse because netgroups are not cached by nscd (but that is a
> story for another day).

        Is this project still needed?  Sparks is planned for the next
        update release of S10.  Shouldn't it handle the performance
        issues noted here?

> 2. Proposal
> -----------

> This is deliberately a file on the local host rather than something
> held in LDAP against the user or host.  There are other means for
> restricting authentication when pam_ldap is used, these based on the
> capabilities of the LDAP server and are sometimes specific to a given
> LDAP server.

        I'm confused with what point is being made.  LDAP is a name
        service and can serve up name service requests such as
        getpwent(), getnetrent(), ...  pam_ldap(5) is a PAM service
        module that coerces an LDAP style directory service into
        an account authority.
        But you know that, thus my confusion.

> There is also purposely no admin command, the customers requesting
> this functionality want a raw file as that is what they already use
> on Linux or on Solaris with the existing open source module.

        IMO, this is an excuse for not providing a properly auditable
        administrative interface and not a reason.

> 4. Examples
> -----------

        See above re man page.

Gary..

Reply via email to