Peter Kirk writes:As I understood it, stability of normalisation (as used as justification for a refusal to correct errors in combining class allocations) is understood as a guarantee that a process will receive a string from another process, or from a backing store, in the same normalisation form that in which it was sent or stored, without the need for renormalisation. But that guarantee fails if the communication or storage system is permitted to perform canonical transformations on the data e.g. for compression.
If conformance clause C10 is taken to be operable at all levels, this makes a nonsense of the concept of normalisation stability within databases etc.
I don't think that the stability of normalization influence this: as long as
there's a guarantee of being able to restore any desired normalized form,
processes can use any canonically equivalent representation of strings.
If stability of normalisation means only that the required normalisation form may be regenerated, and that data which has to be in a particular normalisation form must always be checked and if necessary renormalised, then the main argument for not changing combining classes disappears.
But only if it is able to be sure that this particular normalisation form is stable, and cannot be transformed by a low level process hidden from it into a different normalisation form or a canonically equivalent but not normalised form.In fact, this stability is a great benefit as it effectively gives the full freedom to transform strings into canonically equivalent forms if this is needed: if a database must be built for performance reasons with a particular normalization form, its interface will then be able to perform this normalization freely. ...
... The same is true for data compression algorithms.
So it's the absence of stability which would make impossible this
rearrangement of normalization forms...
Canonical equivalence is unaffected if combining classes are rearranged, though not if they are split or joined. It is only the normalised forms of strings which may be changed. So this is no argument against rearranging combining classes.
-- Peter Kirk [EMAIL PROTECTED] (personal) [EMAIL PROTECTED] (work) http://www.qaya.org/