One of the assumptions of the current implementation seems to be that it is expensive to check for
IObjectToIsolate on traverse. So, collective.siteisolation looks for objects with that marker only once, in the root folder, and memoizes the result.
Is it so expensive to check interfaces during the traverse? Aren't those objects loaded anyway? If not, this could be changed to check all the way through the traverse, and the subsite problem could be solved.
One of the assumptions of the current implementation seems to be that it is expensive to check for IObjectToIsolate on traverse. So, collective.siteisolation looks for objects with that marker only once, in the root folder, and memoizes the result.
Is it so expensive to check interfaces during the traverse? Aren't those objects loaded anyway? If not, this could be changed to check all the way through the traverse, and the subsite problem could be solved.