'SDE' should have fewer privileges than a data creator (in general, and with respect to
the 'dpstr_gis' user's tables), so this would expose the instance to potential corruption
without providing any benefit.
I'm having this issue also. It's so much fun trying to guess what is wrong when the error message doesn't help you at all. The most annoying thing is that it would take five seconds from ESRI to simply add the failing table name to the message and it would save me hours of tedious work.
That "five seconds" estimate supposes that the database has provided this information in the integer error code made available to developers. I'd put the coding time to make 'sdetable -o create_view' easy to use on the order of 8-10 months (especially considering the potential differences of a dozen different RDBMS providers).
There is a better way to create views: Use SQL in the RDBMS directly. Once you've got the view working as desired you can then register it with ArcSDE via 'sdelayer -o register'. This is best practice, and I strongly recommend it.