[TNG] Status (and merging)

Peter Samuelson peter at cadcamlab.org
Tue Jul 25 09:24:42 GMT 2000


  [David Bannon]
> > Seth, is the dual 2.0.7 / TNG model valid (on the one box) ? I was
> > under the impression HEAD/TNG model was suggested once but since
> > March this year Lars FAQ says it does not work.

[Seth Vidal <skvidal at phy.duke.edu>]
> thats suprising.  I know 2.0.7 can join TNG/use it as a password
> server so I can't imagine why they would conflict.

What David was referring to was the work Luke L did late last year, not
long after creating the split-daemon architecture in TNG.  You have
*one* instance of Samba where the HEAD branch does the fileserving and
the TNG branch does the DC and auth stuff.  They communicate through
local sockets, just like the multiple TNG daemons do.  HEAD would first
try a socket and, if nobody was home, revert to its builtin code.

At the time, Luke was very excited about the ability to develop all the
daemonettes independently -- the theory was that development cycles
would be independent and thus short.  (Three weeks, I think he said at
one point.)  And many people did use "mixed-mode", since HEAD
file/print-serving was so much better.

It was a nice theory.  But apparently the socket protocol was not as
nicely abstracted as one might have hoped, and it was not long (two
months?) before HEAD and TNG could no longer talk to each other,
rendering the exercise largely moot.  (The exercise of including HEAD,
that is -- I still think the split daemons in TNG are useful.)

Peter


More information about the samba-ntdom mailing list