Scott Mansfield wrote: > Not sure if we can toss Nettle... NOW I remember why I chose to go > with Nettle in the first place. For the classes in the network kit > that use nettle, they have an inlined method called "GetImpl()" that > returns a pointer to the class' member data, which happens to be a > Nettle object (NetAddress == NLAddress, NetBuffer == NLPacket, > NetEndpoint == NLEndpoint). How would you propose this be handled? > Seems to me that the only way to be binary compatible is to use Nettle. I don't think we should support backward compatibility here. Who ever call BNet*::GetImpl() methods of the BeOS Net Kit C++ classes ??? Dropping Nettle-wrapping compatibility is not required, I guess. If you have give a look (and I hope you did, as Network team leader ;-) ) at current uncomplete but yet not totally empty that the GetImpl() methods, which we *have* to implement as they're public, just return NULL: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/open-beos/current/ headers/os/net/NetAddress.h?rev=1.1&content-type=text/vnd.viewcvs- markup "No Nettle backward compatibility". Or did I miss something here? In this case, please, I *want* to know what!!! Maybe we should switch this thread to the net team ML (openbeosnetteam@xxxxxxxxxxxxx), too ? > Although, I suppose I could code up some Nettle-like construct > (NLXXX-like classes but just the data), and return these from the > GetImpl() method but that also would more than likely violate the prime > directive (R5 binary compatible). -Philippe Houdoin -- Fortune Cookie Says: On Monday mornings I am dedicated to the proposition that all men are created jerks. -- H. Allen Smith, "Let the Crabgrass Grow"