neme resolution

[Jens B. Jorgensen]

You're just getting started. Run a network trace sometime and wonder at the
"extra" netbt-ns operations that go on which I guess have to do with, I'm
guessing, browse master stuff or something like that. Also, you'll discover
that Win95 boxes only answer packets which have a source port of 137. This is
extremely annoying since in unix you have to be root to use ports <= 1024.
Also this means you can't have more than one program listening on this port
and expect things to work normally. Fun! But at any rate, you're right,
windows doesn't do any special encoding to look up NETBIOS names in DNS,
although it's pretty easy to see that it should to work with all allowable
NETBIOS names.

iddwb wrote:

> I was perusing rfc1001 and 1002 last night and noticed something that
> doesn't seem to fit with the way microsoft smb servers find names.
>
> First, the netbios name is supposed to be encoded cause per the IBM Lan
> Technical Ref regarding netbios a name can be any 16 byte pattern -- non
> printable characters included.  I've never seen microsoft make this
> recommendation for dns resolution.  So, how does a ms smb server find a
> netbios name in dns?
>
> Second, since a netname will contain space padding up to the 15 byte, and
> then a hex code for machine name, group name, workgroup name, etc., and
> since ms is not encoding these names, what does the smb server actually
> ask for from dns?
>
> Third, since netbios names can by any 16 bytes, and the reversible half
> ascii encoding method specified in the rfc would imply case sensitivity in
> names, what does the smb server do with uppper/lower case in name
> resolution?
>
> Fourth, since a period '.' is not considered a valid character in a
> netbios name (the original list of invalid characters included things like
> "*", "?", "/", "\" "'"), and since it is possible the connect to a dotted
> decimal notated netname via \\129.219.15.22\sharename, the requester must
> be converting this ip address to a netbios name? or resolving it some
> other way.  Can anyone describe the mechanism used here?
>
> Fifth, while have haven't read the CIFS spec completely, I haven't seen
> anywhere that in CIFS that changes the nature of netbios naming.  For
> example, with CIFS can you partition the netbios name space like a domain
> name space?
>
> David Bear
> College of Public Programs/ASU

--
Jens B. Jorgensen
jens.jorgensen at cmgisolutions.com