Can't interpret GD&T callout

According to ASME Y 14.5, a datum is defined as 
"datum: a theoretically exact point, axis, line, plane, or combination thereof derived from the theoretical datum feature simulator."

So although it's not clear, B Datum in this instance is an implied line between the centroids of the two circles, with allowable feature size MMB bonus for each datum feature.

It's sketchy as a datum, as the origin point is ambiguous... But if taken literally (with the line as ONE feature of size, (Note in the definition it's single tense, not plural)), the origin would be the MIDPOINT of that line.

Is that the intent of the engineer? Unlikely.  I would follow MattR 's suggestion that you use the top left hole as origin and rotate about the bottom right, then trig out the basic rotation offset angle between the two datum holes within that alignment structure.  This would square up your alignment to be orthogonal to the print, to get your basics to report appropriately. 

--If I got any OOS without the B datum MMB shift allowed in all the other non-datumed holes, I would change that datum alignment to a bestfit alignment of the two datum holes, as the alignment.

When my buddy went through engineer school recently, he said they spent about 1/2 of one class period discussing GD&T.  Which makes sense based on some of the drawings new engineers come up with 

When my buddy went through engineer school recently, he said they spent about 1/2 of one class period discussing GD&T.  Which makes sense based on some of the drawings new engineers come up with 

Yeah ASU engineering undergrad program provides less than one week of curricula related to GD&T.  It's appalling.