Association elements are essential for referenceless associations, and can
be very useful for creating links between existing objects for associations
with references also (the latter should be optional controlled by a new
tag).
They are not mentioned at all in section 4 (Schema Design Principles)
despite appearing as Rule Set 10 in 5.2.1. and Rule Set 7 in 5.3.1.

Also the schema and document generation rules for Association elements are
unnecessarily heavyweight and inconsistent with the treatment of references
on classes. For example it does not make sense for an AssociationEnd element
to have the fixed xmi attributes; and it should be possible to dispense with
them completely: letting an Association element refer to the linked elements
using XML attributes. e.g. for association A with ends b and c, a link could
be represented as follows (b1 and c1 are xmi.id's): <A b=b1 c=c1 />).

4.5.3: It would be useful to have an extenderVersion to match
exporterVersion.

It is not clear why a separate XML construct has not been used for proxies
(as described in 4.9) - rather than leaving an importer to check for the
presence of link attributes. And providing no Schema-level enforceable
constraint that proxies have no content (and vice versa).
It would also help to have a stronger statement with respect to the use of
XMI attributes as a 'cache' within proxies: to what extent must they be
accurate and what options does an importer have if they are not? Does
validation include checking the accuracy of cached attributes?
4.6.2: Could Schema mechanisms be used to enforce the constraint that only
one of href and idref is to be used?