1.b4 e5 2.Bb2 f6 3.e4?! Bxb4 4.Bc4
I remember coming across this unlikely gambit in an old David Levy monograph on the Sokolsky Opening, which I used to play as a teenager. My reaction, at the time, was bafflement: "how did they think that one up?". I was even more surprised to find a player called Fischer on the White side of the gambit, which I therefore assumed to be some other Fischer, of whom I had not heard, until I discovered that it was a simultaneous game played during Bobby's famous US tour of 1963-4. (He played his famous Evans Gambit offhand game against Fine not long before - perhaps that's where he got the idea of trying out this line.)
I still confess myself surprised, though - surprised that Fischer played it, surprised that anybody ever played it. Surprised still more to discover, while writing this piece, that Tartakower turned over both Reti and Colle with the gambit. Nevertheless, I still think it's a load of old bosh, though I've never actually played either side of it: as White I always bottled out with 3.b5, and as Black I normally go 2...Bxb4, if you can talk about "normally" where such a rare opening is concerned.
What's also surprising about the sequence is that it involves the development of both White bishops (indeed, of three bishops in all) before any knights have come out. It's startling, when you think about it, how rarely this is the case, for either side. In fact I can't, even after some thought, think of a totally respectable variation where this is the case.
There are various Modern Defences where the Black bishops go to g7 and g4 - selecting one particular move-order in that opening is impossible, but as White I might very well face a sequence like 1.Nf3 d6 2.c4 g6 3.Nc3 Bg7 4.d4 Bg4
which has been played by rather more grandmasters than the aforementioned Sokolsky gambit.
Other sequences for developing the bishops before the knights include, for Black, in the English Opening, 1.d4 e6 2.c4 b6 3.e4 Bb7 4.Nc3 Bb4:
which is fine for Black, although White has not chosen the best line - and for White, a line in the Nimzowitsch-Larsen,1.b3 c5 2.Bb2 d5 3.e3 Nc6 4.Bb5
which is also fine for the side deploying the bishops, but which may not be the other side's best choice.
Which, if it says anything, perhaps says that there may well have been something in the old chess proverb "knights before bishops": even if it be only that its opposite, "bishops before knights", very rarely applies. One wonders why. Because they're targets, when developed too early in the game? Or because knights both attack and defend, more easily, pawns (and squares) in the middle of the board?