I would like to change the format of stabilizer chains for permutation
groups slightly. Because they are described in the manual this would
violate our claim that we will remain compatible with what is described
in the manual. The purpose of this message is to find out whether nobody
is relying on the format of stabilizer chains so that I can change it for
GAP 3.3, or whether I shall only make an intermediate change for GAP 3.3
and wait until GAP 3.4 with the full change.

I want to make two changes, of which only the first would be visible.

Currently the group itself is the top level stabilizer, e.g., the group
record itself has the components '<G>.orbit' and '<G>.stabilizer'. I
want to change this so that the group record contains only the component
'<G>.stabchain', and it is this record that contains the components
'<G>.orbit' and '<G>.stabilizer' for the top level. The idea behind this
is that all components in group record should be independent, so that a
user might delete those whithout the possibility to create inconsistent
group records (as would now happen if the user deleted only
'<G>.stabilizer'). This change is also needed for the second change.

The second change would be to change the base change function so that it
never changes stabilizer chains (destructively), but always creates new
stabilizer chains. The idea behind this change is that quite a number of
functions must copy entire stabilizer chains, because that would fail if
the stabilizer chain was changed. For example 'Stabilizer' (of a single
point or list of points), makes a base change, so that the new base
starts with the points to be fixed, and then installs a *copy* of the
remainder of the chain as chain of the stabilizer. This copying could be
avoided if stabilizer chain would not be changed.

At the same time I will also change 'MakeStabChain' so that it will not
install the stabilizer chain into the group record, until it is complete.
The reason is that quite a few users have complained that after
interrupting a stabilizer chain computation, the group record was in an
inconsistent state (which will cause incorrect answers for 'Size',
'Centralizer', etc.).

Martin.

-- .- .-. - .. -.  .-.. --- ...- . ...  .- -. -. .. -.- .-
Martin Sch"onert,   Martin.Schoenert@Math.RWTH-Aachen.DE,  +49 241 804551
Lehrstuhl D f"ur Mathematik, Templergraben 64, RWTH, D 51 Aachen, Germany