On 28.08.2009, at 00:08, Ken Causey wrote:

> Now that's an interesting suggestion that had not occurred to me.  Is
> there some way we could leverage this fact to avoid this sort of  
> problem
> in the future, it seems to have come up more than once already.
>
> On the one hand perhaps the updating logic could create a new  
> temporary
> project and enter it before starting the update then exit and delete  
> it
> when the update is complete?

Maybe, but it smells like an evil hack.

>  Alternatively could the logic do some
> subset of what happens when a new project is entered but as effective
> for this purpose?

Possibly, but it'd be an even more devilish hack.

>  On the third hand does atomic loading address this
> problem more effectively and more directly?

I'd say so, yes.

- Bert -