> Fixing this is more difficult, because there is no common base to

all

> the different components that use AutoLabel. What I propose to do

is

> prevent assigning the model property of a component that is already
> being used as a model. This is not perfect...
>
> You could assign the model property of ControlA to ControlB. Then

you

> can assign the model property of ControlC to ControlA. Hence

ControlA

> gets both a model property assignment and is used as a model itself.
> However the references are still not circular, and it will remain
> impossible to assign the model property of ControlB to any other
> control - because ControlB already has a model reference from
> ControlA. (Clear as mud ;-)

Ok, it's not so simple, but let's figure out what's the real use of
the property. I think I will setup on my form some "model" autolabel
and make all the other behave like those. I don't think having
circular or chained model assignment is very usefull. So you could
have that you can assign model property a control that has not model
property assigned itself. You can have ControlA and ControlB as
models, and ControlC,D,E have ControlA as a model, while ControlF,G,H
will have ControlB. This is a sane aproach, otherwise you will get
troubles soon (tracking all the web of model assignments).

>
> So the questions are...
>
> 1. Does this sound acceptable (am I missing anything in my logic)?
>
> 2. Should the rejection of a model assignment result in a message,

or

> do I simply ignore the attempt?

Maybe a message will save you from plenty of "this property does not
work" e-mails ;)

>
>
> > > c) I want to increase the space between the end of the label
>
> I explained how to do this. Perhaps there are still some issues in
> regard to property names or property availability but I dont want to
> tackle those right now.

I tell you again that you have had a wonderful idea with your
autolabel, thanks. I think that you should fix Margin behaviour as
soon as possible, otherwise a lot of code will be done with a wrong
property behaviour... and you will not fix that for "compatibility"
with the past.
See my post of 2001/02/22 and the one of Leonardo Quijano, please.
Thanks a lot
Marco Menardi

>
>
> Geoff Worboys
> Telesis Computing