cshring at gmail.com
Wed Jun 21 07:34:08 GMT 2006
On 6/8/06, Gerald (Jerry) Carter <jerry at samba.org> wrote:
> I would definitely support this effort. I've wanted it for a while.
> My preference, rather than to backport libnet code from Samba 4,
> is to provide an API in stable library that is a thin layer wrapped
> around the underlying rpc_client infrastructure. That way we could
> start using the IDL from Samba 4 in the future without breaking
> existing applications. This allows us to (a) get a working library
> out there faster using the existing Samba 3 releases as a vehicle,
> and (b) gives us a path to migrate to the improved RPC support
> developed in Samba 4. I also think that porting the Samba 4
> IDL should be considered part of this project.
> We already have the beginnings of this in the libmsrpc.so library
> in SAMBA_3_0. I think it is important to make this RPC interface
> available to other applications besides Samba (e.g. Mono) as soon
> as possible.
> Some of the management features you outline don't have a direct
> mapping to existing RPC pipes so we need to think about what to
> do with those.
Following up the "thin wrapper api" idea a bit more - is msrpc the
ideal place for such abstraction or some other code-subtree common to
Samba3 and 4 ?
The thin wrapper idea is appealing since it can connect to newer and
newer api thrown up as part of Samba3/4.
Case in point : James mentioned the need for the performance
monitoring api to be lightweight. The wrapper can instead sample data
via xmetrics api providing it to the consumer. The wrapper obviates
the need for James to provide his api as part of libnet or anysuch.
Backporting libnet is tanamount to having Samba4 finally in place of Samba3. :)
I am willing to take it up, however I need suggestions / pointers to
avoid painful re-work later on.
More information about the samba-technical