I agree to a large extent with the excellent summary of the situation
of Root under NT by Gordon. I have some comments:

 -  Root under NT is already multithreaded. In fact, Root could
not run if it was not. There are, however, still many problems
with the thread implementation as it is now. In many places,
there is no check for synchronization. CINT itself must be run
in the main thread.

 - Implementing the class TGWin32 is an urgent issue. The new GUI
was a big step on Unix. Currently the GUI on NT is implemented
by something like 30 classes. The browser has a different behaviour
than on Unix. Having TGWin32 will be a considerable simplification
on the maintenance procedure. We currently cannot use the new GUI
to replace several existing widgets (DrawPanel, Fitpanel,
Attribute panels, etc).

 - We have defined one single abstract base class TGXW to provide
all the functions for the GUI and the basic graphics. The choice
of the API was certainly driven by our experience with X11. We will
probably have to make some slight changes in TGXW during the
prototyping of TGWin32.

 - Concerning the interoperability of Root under NT with the
NT framework in general (Excel, Word, PowerPoint,etc), we had
the pleasure to see the announcement last week of ActiveRoot
from Andy Haas. ActiveRoot is a small front-end to Root that
enables the import of Root objects into all the Microsoft products.
For more details, see the URL:
 http://glast.phys.washington.edu/~haas/winroot/index.html

 - Porting under VisualC++ version 6. This work is now completed
thanks to Sverre Jarp. During this exercise, we found one
compiler bug in VC++ reported to Microsoft. Meanwhile, we have
a workaround.

Rene Brun