Must a Samba PDC use encrypted passwords?

Albert Chin-A-Young china at pprd.abbott.com
Fri Jun 5 21:20:20 GMT 1998


On Sat, 6 Jun 1998, Gerald Carter wrote:

>Albert Chin-A-Young wrote:
>> 
>> If you're running NIS, I'd like to see a modified rpc.yppasswdd that
>> updated NIS passwd and NIS smbpasswd at the same time. Samba would
>> then have to be modified to use NIS for smbpasswd. I'd also like to
>> see Samba refer to both smbpasswd and passwd (if the login name is not
>> in smbpasswd, then look in passwd). The latter is important until
>> everyone changes their password.
>
>making the smbpasswd file available via NIS IHMO is a bad idea. 
>Remember that the hashes in smbpasswd are plaintext equivalents.  Better
>to use the security = domain model and authenticate against one server
>that has the smbpasswd file local.

This depends on your implementation of NIS. The way we have things
here, every machine is an NIS master and, therefore, can only bind
to itself (we could also make the Samba servers members of their
own domain). If smbpasswd were in NIS, then two Samba servers,
or more, could share the same file and you could have updates
occur centrally with a modified passwd/rpc.yppasswd combination
(it's also easy enough to distribute smbpasswd to only a few NIS
servers). We distribute NIS maps as flat files and could easily
rdist them with ssh to severely decrease the security flaws in the
idea. I don't care for single points of failure but if 'password
server' supports more than one password server, then I'm all for it
(but then you still have the problem of keeping smbpasswd in sync).

>                            Gerald ( Jerry ) Carter	

-- 
albert chin (china at pprd.abbott.com)



More information about the samba-ntdom mailing list