cubicweb #3891985 buggy 3.19 composite removal? [open]

changeset be001628edad which introduced discussed semantic change of composite deletion introduce IMO potential problem by removing the operation. I'm very fine with the discussed semantic change though the changeset introduce undiscussed bits. Following the rule of the least surprise I would expect that in a same transaction:

  • I can explicitly remove the composite parent then its children
  • I can explicitly drop a composite then reparent its children

Now in both case the children will already have been removed leaving me no choice.

Rational for this part of the change has not been explained at all nor in the commit message nor in the changelog.

done in<not specified>
closed by<not specified>