Am 14.07.2006 um 16:13 schrieb Vlad Seryakov:
I still think that you could build ns_callback command that will be
similar to ns_conn, i.e. it works only in the context of callback like
ns_conn in the connection thread only. This way you can extend it
as you
wish and be backward compatible
Almost! This is what I also thought at the first glance!
But when you think deeper...
My idea was to add more information available to the callback
in particular, the events of adding, calling and removing.
BUT, in order to do that you MUST actually call the callback
once at the registration, regulary during the execution and
once at the de-registration. Now, this can't be easily done
as callbacks may have never been designed for such call
pattern! Therefore it is not possible to automate this task
as I wanted it.
If the callbacks are written with this notion (i.e. that
they will be called at least 2 times (on registration and
de-registration) then it is another story. But since we
cannot really know how all this will react, we can't do that.
Effectively I can build a kind-of ns_callback command which
is valid only in the context of the callback. But for it to
be used the callback must be executed(!) and this is where
all comes down. Understand?
That's why I had to throw this idea away... I gues I was just
not thinking enough beforemath (how could I, we have 32 grades
centigrade here and we're all sweating our brains out...)
Cheers,
Zoran
