On Sat, Jun 28, 2003 at 11:10:45PM -0600, [EMAIL PROTECTED] wrote: > Hello! > > I am using qmail-ldap in two diferent setups: > > a) with a virtual-user setup, where all accounts are mapped to on > uid/gid via ldapuid/ldapgid. > > b) ldap is used to maintain the unix-account attributes of all accounts > on a whole university campus: each uid/gid is different. > The nis.schema object is used with the openldap - slapd server. > > In the second setup I found, that it was not sufficient specifying the > uid/gid of each user account, they had to be replicated in the qmailUID > and qmailGID attributes provided by the qmail.schema. > > While I find it probably useful to be able to specify different > qmailUID/GID's then the posix-account uid/gid's, I was expecting, that > qmail-ldap would fall back on the latter one, if qmailUID/GID was not > specified in the ldap-directory. > > This also hit me in the first setup, when I created some "standard" > unix-accounts with different uid/gid's and wanted mail delivered to them > via qmail-ldap. > > Is this a feature? And what I am looking over in this case? >
You can change the ldap attributes the qmail-ldap uses at compile time, just edit qmail-ldap.h. > If not: could posix-uid/gid fallback be implemented - please? :) > If we will add a fallback to the nis schema is currently unclear. First I like to have mailGroups, smtp-auth and a deliveryMode cleanup. > The QLDAPTODO has one entry which seems to come close to this issue: > > "- test what happens if only uid and mail is defiend in ldap or similar uncommon > cases (seems to be no problem, or?)" Close but not close enough. This is more of a regression test to see what is the minimal ldap entry that makes sense and what happens if less was specified. Currently the minimal ldap entry needs to have mail, uid and mailforwardingaddress. -- :wq Claudio
