The same when you have a drop-down. It starts with blank (if you don't define a default). Once you assigned a value from the drop-down there is no way back.
Well not to go all Paolo on you here Q, but IMO that's actually right. An enumeration type is one-of-a-set, it's not nullable. And that's consistent with the behaviour of Dates and most other structured tag types (except RefGUID, which you can set to <none> in the selector dialog).
Unless the standard actually prohibits it, I think it should be possible to specify that a structured tagged value should be nullable.
In the structured tag value definition this could be implemented with a new reserved word Nullable=True|False;, with the tagged value connector it's a question of whether the multiplicity includes 0.
If not nullable, current behaviour; if nullable, you would be able to clear the tagged value.
Certainly, EA doesn't have any problems with tagged values actually being empty, so this shouldn't be impossibly hard to do.
/Uffe
A couple of issues here. Firstly, Uffe is correct in saying that an enumeration type is notionally not nullable. But where data entry is concerned, there should
always be the possibility of extrinsic
[1] values to handle data entry concepts (uninitialised, missing, removed etc.). This can be accomplished for most tag value types (and their equivalent datatypes).
Dates aren't an enumeration set, because they have a range. But that having been said, it is still necessary to handle extrinsic values. Where the dates are stored as strings (such as in EA tags), then string captions for the extrinsics can be allowed. Where the dates are not stored as strings (even numeric strings), then specific dates can be used to represent the extrinsics. Over the decades, I've standardised these values and they seem to work quite well. I can share if asked.
As to whether the Standard allows Nulls... If it doesn't, then Sparx is in violation (see the discussion regarding Tag default values). So we should assume that it doesn't explicitly forbid them.
As a Data Architect, I abhor NULLS, but if you really insist on allowing them, then it should be explicit as suggested above.
Paolo
[1] Extrinsic values are, by definition, not nulls.