samba - May 1998 - The answer is blowing in the WINS

Dear Ladies and Gentlemen,

I am writing to ask your opinion on how many WINS and WINS proxies are
necessary in a routed network.

The questions are in the context of using the Samba 1.9.18 nmbd as a WINS
or WINS proxy with MS Windows for Workgroups or MS Windows 95 clients.

Is it

A) 1 WINS, 0 WINS proxies (every host asks the WINS for NetBIOS names with
Unicast)

B) 1 WINS, n WINS proxies (where "n" is the number of routed subnets,
and
hosts ask the local NMBD with Unicast, and this NMBD, if necessary, asks
the WINS)

(B) seems to be wrong since NMBD 1.9.18p7 running as a WINS proxy reports
messages like

process_name_refresh_request: unicast name registration request received
for name 03151<00> from IP 10.0.3.251 on subnet UNICAST_SUBNET. Error -
should be sent to WINS server

However if that is the case, what is the use of the WINS PROXY
configuration option ?

Why does the samba documentation suggest that in a mixed NT and Samba
server environment, an NT server should be the WINS ?

It seems to me to be much easier to manage a Samba WINS than an NT WINS
since

. the Samaba WINS can be managed remotely by telnet
. the Samba WINS database is in text form and so can be easily published on
the WWW with CGI (I prefer the ISC DHCP server to the NT one for this
reason also)

The MS databases are kept I think in "jet" database format.

. the Samba logging is much much better
. the NT WINS database replication is of *no* value because unlike a DNS
zone file the data in a WINS database is volatile.

and
. Unix hosts are so much quicker to respond to my input than any MS
product
. There are more cheap, easy to use, readily available tools for Unix than
MS products (Perl being the only exception).

I'd really like to hear your answers to these questions.

Thank you,

Yours sincerely,


S Hopcroft