[Fwd: C++]

Jelmer Feenstra spark at dds.nl
Tue Oct 10 20:38:40 GMT 2000

On Tuesday 10 October 2000 22:30, gcarter at valinux.com wrote:
> Jelmer Feenstra wrote:
> > And the current libsmb project is rather dead, we
> > actually need a good implementation of the
> > clientside SMB functionality wrapped up in a nice
> > library (preferably C++ I guess). However, talking
> > about this has happened for about a few weeks now and
> > nothing was decided on this matter (for I can tell).
> Talking about Samba being implemented in C++?  That
> won't happen unless all developers concurrently decide
> to rewrite 200,000+ LOC from C to C++.  No big win.
> Don't get me wrong.  I like C++.  Would be my preference
> for a new project probably.
> The bottom line in that a conversion to C++ would not be
> cost effective given the current investment in the code base.
> No one can guarentee that C++ would make Samba more managemable,
> portable, than it already is.  SMB is the ugly part. :-)
> If however, you were talking about a C++ based smb library,
> then what is the question again?

My question is actually, what is the best path to walk if I would like to 
write an application which would need fast access to smb; reading/writing 
files, browsing hosts etc ?

Libsmb++ would be a nice solution, however the project is rather dead at the 
moment, and I don't really think it will be revived by the author, Nicolas 
Brodu (as he is no longer on a windows/smb based network).

So I asked (in another email) if it would be possible to make smb a lot more 
practical (on linux systems in general) by using a sort of userfs 
(kernelspace) in combination with a daemon (userspace) that provides the 
needed file/listing/browsing operations. This way it would be possible to 
have a sort of /proc filesystem which would actually represent the current 
status of hosts on your network, allowing you to simply changedir into them 
and use the files that are there. Autofs in combination with smbmount is not 
sufficient (no browsing of shares/computers) and in my opinion it's not the 
right way to go.

If this can't be done, I think a C++ based smb library would be very usefull 
(KDE2 comes to my mind, which is currently using the not-finished libsmb++ by 
Nicolas Brodu).


Jelmer Feenstra

More information about the samba-technical mailing list