multiple nmbds domains workgroups

Luke Kenneth Casson Leighton lkcl at
Tue Sep 18 04:07:02 GMT 2001

On Mon, Sep 17, 2001 at 06:33:36PM -0700, Andrew Tridgell wrote:
> I've just looked over our discussion on this from the time (from
> 1996!) and the idea of tying the workgroup to a netbios alias was
> indeed raised - I'd forgotten about that solution. 
> It looks like we didn't have a really good solution to the *SMBSERVER
> and NBT node status problems (remember MS was pushing the horrible
> idea finding the netbios name for an IP by looking for the first type
> 0x20 name in a node status response on name '*'). 

mmm :)

> I suspect these
> problems wouldn't be major show stoppers in practice because the
> browse synchronisation process is unlikely to trigger such atrocities.
> It's probably not worth pursuing this now with the port 445 change and
> lack of the session request, but we probably could have made it work
> OK at the time.
yeh, what the heck :)

thought about this some more overnight.

if you can respond to *SMBSERVER (via an smb.conf option)
with a NetBIOS Session Reject message, then the client (win)
will do a Node Status lookup automatically, and connect
to the real NetBIOS name from there.

then, create an smb.conf suite including by called name
config file = /usr/local/samba/etc/smb.conf.%m
[config dir] = /usr/local/samba/%m/

[not sure what the config dir option is really]

then, in nmbd on every UDP mailslot received, set
the remote_machine global variable from the UDP
calling name, and call reload_services() *before*
responding to the UDP mailslot.

this'll probably be horribly slow, but it would
work with minimal coding.


More information about the samba-technical mailing list