[jcifs] (no subject)

[Michael B Allen]

>> Actually there is another *very easy* solution to this problem. The
>> NetBIOS 0x1C group name lookup returns a list of domain controllers. As
>> it
>> is we just read the first and return that.
>
> Note that the first is always the PDC.  The rest are BDCs.  I *think* that
> the BDC entries are ordered with the most recent refresh at the top of the
> list.

I noticed in the list of ~15 address for my domain at work that the first
and third were the same. Mmm.

>
>> I have modified the
>> NameServicePacket.readResourceRecordWireFormat() method to decode more
>> of
>> the list depending on the value of a static 'addrIndex' member. IOW, if
>> a
>> NetBIOS name lookup returns more than one name, the code will pick the
>> next address in the list or reset addrIndex to 0 if it reached the end.
>
> Um... "if a NetBIOS name lookcup returns more than one" IP address.
>
> If I read this correctly, the method will, if called repeatedly, enumerate
> the IPs in the list, yes?  Checking addrIndex will let you know when
> you're at the end of the list.  Have I got that straight.

Right. The addrIndex member starts out at 0. With each query it returns
the address at that index but not before incrementing addrIndex. When a
query enumerates the entire list before reaching addrIndex, addrIndex is
reset to 0 and the last address is returned.

But it sounds like maybe we should also limit the value of addrIndex so
that only the first N addresses are ever used?

>> Now I just have to modify the NtlmHttpFilter to do an
>> NbtAddress.getByName( domain, 0x1C, null ) lookup instead of the usual
>> 0x20 lookup.
>
> If you do a broadcast query, you will only get local LAN DCs.  That's
> okay.  It's just that you'll get zero or more packets instead of a single
> response (with multiple IPs) from an NBNS.

Well I don't think the user is concerned with load balancing between
domain controllers if they don't bother to set WINS.

>
>> The effect will be that a different domain controller will be
>> returned with each NTLM http negotiation and sessions will be
>> distributed
>> evenly across them. The change is about 5 lines of code so unless this
>> trips up Chris' NetBIOS sensors I think 0.7.15 is coming up ...
>
> Only the caveat about broadcast vs. NBNS queries.  As long as there's a
> way to specifically retrieve the first entry if there are multiple
> entries, I think this'll work.

Again, I don't think we should be concerned with broadcast but the
behavior will not change if the user does not set WINS.

Mike

-- 
A program should be written to  model the concepts of the task it
performs rather than the physical world or a process because this
maximizes the  potential for it  to be applied  to tasks that are
conceptually similar and, more  important, to tasks that have not
yet been conceived.