I'm investigating deploying multiple samba servers to remote offices which have slow links to central LAN with NT4 as PDC, hoping to provide the following :- 1) Local samba fileserver in each office for faster access to shares, which can be rsynch'ed or backed up across the slow links to the central LAN overnight. So far so good :-) 2) Impove logon speed for for users in each office by reducing netlogon share + profile share + SAM traffic over slow links, hmmm, this would normally be the job of an NT BDC but samba 2.x can't do that and it's probably not acceptable to use samba as PDC instead of NT4 :-( What if ..... Install samba + winbind + pam_winbind in each remote site Create netlogon share on each samba server and rsync from PDC to reduce logon script traffic over slow links. Create roaming profiles share on each samba server and use %logonserver% to specify user profile path in user manager to avoid profiles being retrieved and saved across slow links. Set "domain logons = yes" in smb.conf so samba "looks like" a BDC to clients in remote sites When clients authenticate to samba it should hand off the authentication to winbind, presumably this will still work despite "domain logons" setting? Will winbind's caching reduce authentication traffic over slow links to PDC? Or doesn't it cache password hashes? Or would the time to check the SAM sequence number is unchanged be about the same anyway? Presumably each samba server would be best as the subnet master browser? What about WINS? if remote clients use their local samba server for WINS they will find the local samba server as their BDC through DOMAIN#1c netbios records, but the multiple remote samba WINS servers can't replicate with each other (or the main LAN's NT4 WINS server) Might it be better to go against the usual grain and let the remote sites use broadcast for netbios name resolution rather than WINS? The samba HOW_TOs concentrate on providing BDC functionality as redundant backup for samba PDC, but providing geographically dispersed BDC functionality is another big reason for using them, has anyone tried the above or anything similar to crack this particular nut?