In an opening booklet of the 1960s, Bent Larsen had recommended 3...d5 4.exd5 Nf6. Later in Kaissiber he preferred 3...Qe7, too. It is surely a good reply, and therefore the other move order (Nf3 first) should be considered.  

kylemeister wrote on 08/21/17 at 19:15:37:
 
That's fine when the two alternatives are of roughly the same value... But what about 1.e4 e5 2.d4 exd4 3.c3 Qe7 4.cxd4 Qxe4+ 5.Be2! - Rudolf Spielmann's move?


When Black expects 5.Be3, giving him a comfortable advantage, but the opponent suddenly isn't cooperative and plays for fast development, what is Black going to do? If he does not take the second pawn, he accepts a balanced position. Taking the second pawn seems obligatory here, for an ambitious Black player, right? (By the way, I see that Kaufman doesn't mention 5.Be2.)

Fllg wrote on 08/21/17 at 18:24:46:
+1

One pawn yes. Two pawns no.

Fllg wrote on 08/21/17 at 18:24:46:


I seem to recall Larry Kaufman not thinking so in his earlier repertoire book.

Martin Lokander's book The Open Games with Black (2015) recommends to take both pawns: 1. e4 e5 2.d4 exd4 3.c3 dxc3 4. Nf3 Nc6 5.Bc4 cxb2!? 6.Bxb2 Bb4+ 7.Nc3 Nf6 8.0-0 Bxc3 9.Bxc3 0-0. The author only considers the reply 10.Re1. White has the stronger 10.Ng5!.


For example 10...h6 11.Nxf7 Rxf7 12.Bxf7+ Kxf7 13.e5 Ne8 14.Qb3+. This position offers White not only full compensation for the pawns, it is also exactly what a gambit player wants. If 14...Kf8, 15.Rae1 threatening Re3-f3+. 

By the way, the similar 9...d6 10.Ng5! had already been played in four games - three wins and one draw for White.