ActiveX and RR

[Chipp Walters]

Dar Scott wrote:

> Many automation COM services do not need a control.  I realize that most 
> folks want the control, but where I could have used this over the past 
> couple years, I have not needed the control.

I agree, this actually would be the best case scenario for an
ActiveX/COM wrapper external.

> The calls should be more straightforward.  The callbacks may be a 
> problem and there may be a need to get across a thread boundary.  In 
> looking at similar problems, I think the Revolution app may need to poll 
> for the callbacks.

You are correct, this is the big hurdle.

> I don't think this is as bad as it sounds.  The first step is to come up 
> with the right mapping of types.  The pointers are known and are not 
> arbitrary so a scheme can be set up for those.  The implementation step 
> might be tedious but half can be done in transcript.

Certainly a naming convention could be setup where variable prefixes in
RR could be used to 'type' data on the other side. But, even then,
pointers referring to pointers still remains a problem... And arrays,
too (Arrays of arrays of arrays...I know I know...'Dar's Boxes!' :-)

> Such a thing would have been handy for some DCOM work over the past 
> couple years, but I don't know how much that might come up again.  I 
> have toyed with the idea of a generic dll interface, but that has crash 
> risks if the developer gets the interface wrong.

No doubt, making a crash proof version would be a challenge. In fact, a
main reason we haven't taken this on is primarily one of support.