[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: back-passwd patch for chfn data



At 11:58 PM 3/5/99 -0800, Andrew Daviel wrote:
>I'm probably doing this all wrong, but then I'm probably
>not going to do much more work on this, and I thought I'd at least share
>my ideas even if my code's all rubbish :-)

You might check out the following ITS#40.  It has a few suggestions
to improve the passwd backend.
  http://www.openldap.org/its/index.cgi?findid=40

>We're setting up a Linux box to run sendmail and IMAP, with a captive
>shell to let people change their display name and password.

I would recommend avoiding system accounts altogether.  Why not
integrate sendmail and IMAP to an LDAP directory.

>The attached patch against ldap/servers/slapd/back-passwd/search.c
>splits out these extra fields from the GECOS, and adds a mail alias
>created by underscore substitution from the fullname field.
>I assigned:
>  cn = real name
>  telephonenumber = work phone
>  homephone = home phone
>  mail = pw_name
>  mail = real_name

mail shouldn't be used for aliases.  mail is multivalued attribute
specifying the e-mail address(es) of this entry.  Most messaging
servers send a copy of the e-mail to EACH of these addresses.

>  l = office
>  givenname = real name[0]
>  sn = real name[n]
>
>There's problems with dealing adequately with all variants
>"Andrew Daviel", "Andrew J Daviel" "Rip van Winkle", "A. Daviel" etc. - 
>it gets RvW's last name as "Winkle" instead of "van Winkle". chfn
>disallows "Daviel, Andrew" which removes one hiccup, at any rate.  

I wouldn't bother with attempting to parse givenname and sn out of
the full name.

>One could perhaps assign "ou" or "office" to office; I arbitrarily
>assigned "l" (= location) since it appears in the Netscape listing (though
>as "City");as a list for primarily internal consumption room numbers
>seem a bit more useful than City....

Locality seems fine here. 

>As regards the code, clearly it needs fixing somewhat if anyone wants to
>adopt it officially; it needs to inherit DOMAIN and MatchGECOS
>properly from config.h or something, and my "C" is a bit inept, though I
>believe it doesn't crash (at least, not straightaway...)  :-7

They should be backend configuration options.

As I noted in ITS#40, enhancing back-passwd would be a good
introductory project for anyone interested in gaining experience
in the slapd code.

Kurt