Seems to be quite usable at this point.

(especially external methods) on instances that are not acquiring.

the user uses the auto-create options on the add Folder form.

has_role.

*unless* they internally reached the word threshold, at which point
they did their own commit and reset their counter.  So, the value
returned is allways less than the threshold.  This is so huge
documents get the benefit of subtransactions, while lots of small
documents still get subtransacted too.

Catalog uses to decide when to commit a subtransaction.  More words:
less commits, faster speed, bigger memory footprint.  Less words: more
commits, slower speed, smaller memory footprint.

ballooning beyong the capacity of the machine.  Note: while mass
indexing, the memory footprint of Zope will get huge.  There is NO way
around this.  incrimentally index if you want your process to stay small.

First cut at a filtered product add list. This will keep objects out of the product add list for which the user does not have adequite permissions. This is implemented with a new method, 'filtered_meta_types' which returns all the meta_types for which a given user has permissions.

Store the permission in meta_types. This facilitates a filtered product add list. Note: control panel products and old style product registration do not add a permission key in the meta_types.

Added a has_permission method to user objects. This facilitates finding out if a user has access to a given object. Used for filtered product add list change.

handled specially.

tuples or lists of dictionaries.  This is somwhat useful for xml-rpc.

words indexed for that call.

will commit to a subtransaction, thus keeping memory consumption
down.

we have to pass this special kind of exception through.

You must *never* assign exc_info() (or sys.exc_info()[2] or
sys.exc_traceback) to a variable unless you also unassign it in a
finally or you'll generate a circular reference in the stack causing
all kinds of interesting things to get leaked.

mind empty tuples or lists.

doesn't mind false bodies.

Did I break anything? :)

Also changed xml-rpc handling so that index_html is not used.