Christopher R. Hertel crh at
Thu May 1 18:58:00 GMT 2003

On Thu, May 01, 2003 at 06:52:08PM +0200, Volker Lendecke wrote:
> On Thu, May 01, 2003 at 09:33:17AM +0200, Stefan (metze) Metzmacher wrote:
> > nmbd read's the wins.tdb on start and then holds the a list in memory.
> > then it flushes all entries to a new wins.tdb file periodicly and delete 
> > the old file!!!
> Let me then ask some provicative question: Why a wins.tdb at all???
> Simply linearly reading and writing a wins.dat can not be that much
> less efficient than a tdb. I even would suspect it's more efficient.
> I thought that this has made to co-operate with winsrepld. Obviously
> this can not be the case.

Volker's right.

The inherent problem (one which I dealt with when I converted nmbd from
using linked lists to using the binary trees) is that nmbd is designed
around the idea that the entries are in memory and always available.  
Pointers are passed around and there is no clear delineation between a
record in use and a record in storage.

We need a re-design and re-write of nmbd to handle this correctly.  I 
started one about four years ago but...  Well, I have a day job and a 

This is probably something that can be addressed as part of Samba 4.

Chris -)-----

"Implementing CIFS - the Common Internet FileSystem" ISBN: 013047116X
Samba Team --     -)-----   Christopher R. Hertel
jCIFS Team --   -)-----   ubiqx development, uninq.
ubiqx Team --     -)-----   crh at
OnLineBook --    -)-----   crh at

More information about the samba-technical mailing list