bug in clsql MOP usage? -- change-MOP-class forbidden

Hi,

it seems a paradox that while CLISP announces its full MOP support, cl-sql does not fully work with CLISP because of MOP issues.

It seems that cl-sql is doing something forbidden by the MOP, which CLISP catches while currently supported implementations don't.

cf clsql-3.0.8/tests/test-init.lisp:
CLSQL-USER[26]> (defclass thing ()
  ((extraterrestrial :initform nil :initarg :extraterrestrial)))
#<STANDARD-CLASS THING>
CLSQL-USER[29]> (def-view-class person (thing)
  ((height :db-kind :base :accessor height :type float
           :initarg :height)
   (married :db-kind :base :accessor married :type boolean
            :initarg :married)
   (birthday :type clsql:wall-time :initarg :birthday)
   (bd-utime :type clsql:universal-time :initarg :bd-utime)
   (hobby :db-kind :virtual :initarg :hobby :initform nil)))

*** - UPDATE-INSTANCE-FOR-DIFFERENT-CLASS: The MOP does not allow changing the
      class of metaobject
       #<CLSQL-SYS::VIEW-CLASS-EFFECTIVE-SLOT-DEFINITION EXTRATERRESTRIAL
         #x2040F18E>

sql/metaclasses.lisp has a call to CHANGE-CLASS
(defmethod compute-effective-slot-definition ((class standard-db-class)
                                              #+kmr-normal-cesd slot-name
                                              direct-slots)
  #+kmr-normal-cesd (declare (ignore slot-name))
  (let ((dsd (car direct-slots)))
    ...
    (let ((esd (call-next-method)))
      ... 
         (let ((type-predicate #+openmcl (slot-value esd 'ccl::type-predicate))\
)
           #-openmcl (declare (ignore type-predicate))
           (change-class esd 'view-class-effective-slot-definition
                         #+allegro :name
                         #+allegro (slot-definition-name dsd))
           #+openmcl (setf (slot-value esd 'ccl::type-predicate)
                           type-predicate))

Here's what Bruno Haible (author of CLISP's MOP) said on this issue:
>Yes. There are four things that are not allowed with 
>slot-definition objects:
>
>  - Calling CHANGE-CLASS on them - explicitly forbidden,
>  - Calling (SETF SLOT-DEFINTIION-xxx) on them - these setters are not
>    specified by the MOP,
>  - Calling (SETF (SLOT-VALUE ... 'xxx)) on them - the slot names are
>    private and not specified by the MOP.
>  - Overriding the generic functions SLOT-DEFINITION-xxx so as 
>to return
>    a different result that you compute from other parts of 
>the slot-definition
>    object - explicitly forbidden.
>
>In other words, once a slot-definition object is created, it's immutable,
>regarding the MOP specified accessors. (You can have mutable 
>slots of your own, of course.)
>
>The only way to influence the creation of slot-definitions is 
>therefore to
>use INITIALIZE-INSTANCE :AROUND methods on a particular subclass of
>STANDARD-SLOT-DEFINITION.

What would be a good way to resolve the issue?
It would be worthwhile if a) cl-sql worked with CLISP and b) it contains correct (or portable) code (if it's broken).

Regards,
	Jörg Höhle.
Tested with cl-sql-3.0.8, CLISP from CVS and my independent implementation of UFFI for CLISP (see clisp patches section on sourceforge).