In order to distinguish timestamp (without precision) and timestamp(6) the without_precision platform option was added.

Why is this distinction necessary? Not only that the type always has precision (whether it's specified explicitly or not), it makes it possible to specify both precision and without_precision on a column (or omit both), which makes no sense.

When the column definition defined by the application and containing without_precision is compared with the one introspected from the database, how should this comparison work?

Why is this distinction necessary?

Because PostgreSQL does it. Column type can be altered from timestamp to timestamp(6) and vice versa (e.g. by schema:diff). So we need a way to save that differecne on introspection. And we can't use precision: null because it is already used for timestamp(0).

So the mapping here:

Maybe that awkwardness was the reason why in #5961 used a scale property instead of presicion - the scale default is zero. I think it will be best to use "scale". What do you think?

Because PostgreSQL does it.

In my understanding, the difference between timestamp and timestamp(6) is purely syntactic. The default precision is 6.

CREATE TABLE test_timestamps (
  ts1 TIMESTAMP,
  ts2 TIMESTAMP(6)
);

SELECT
  column_name,
  datetime_precision
FROM information_schema.columns
WHERE table_name = 'test_timestamps';

The above query will produce 6 for both columns.

Column type can be altered from timestamp to timestamp(6) and vice versa

They shouldn't be considered different. If they are, it's a bug.

Or, more specifically, the default precision of timestamp columns in DBAL is undefined, that's why they may produce a diff. But the point is, this issue needs to be addressed by improving the type system and type comparison logic, not introducing extra type parameters.