> > The 10 MB bloat is only there because we use a static library. > > For R2 we could introduce a C++ stack without having the bloat. > > Currently, it is just a workaround. It will work, but not as > > efficient as it > > could do. > > After R1 it will work very good. Or do you think differently? > > Yes, I do - I don't want to have that bloat in R1, even if it's only > temporary, at least not for the whole stack. Then, the best solution is to implement shared kernel libs for R1. ;) If we had shared libs support in the kernel which language would you prefer for this task? > > It is clear that nobody should ever want to create an instance of the > > manager object, but this is the same as making all methods public and > > saying > > that some methods must not be used because they are internal methods, > > though > > they are defined public. > > If the ppp stuff is clearly separated from the rest, and the > libkernel_ppp.a is reasonably small, I wouldn't mind if you go C++ only > there - just take care of what you do. The templates that I use for the implementation could blow up the code to 200K, possibly. :( Without the templates and void* lists it could be reduced to around 80-100K (these are only expected values because currently nothing compiles). How can you produce 50K of code in two weeks? I needed more than two months to get that far! :) How many hours do you work on the project per week (this is a question to all of you)? > > Probably we have the wrong programming language, but that is another > > topic. > > :) > > I have to disagree: C is not the wrong language, C++ is just better :) Both have the problem that they do not produce good code on their own. ;) > > FSM_info structure even if you should not do that. > > #define private public > ;-) I will put it into cpp.h. :) Waldemar