An attacking game... in the french? By black?

Here is a nice example of how well black can liberate himself if white plays the french carelessly.

That's correct. Winning the pawn both wins a pawn, and critically weakens white's dark squares. Without that, black might not have quite enough.

And I think I change my mind about ...f6 being the best move, after looking at the game explorer (on black's 9th move there were a few master games played), and now ...a4! looks best, taking b3 from the knight first. It seems that after I played ...f6 white should have, after exf6, Nb3! when after ...Bd6 white can get in Be3 with decent control over the central dark squares. If black just plays ...e5, I think (I'm too lazy to put this on a board right now!), white can play Nxc5 Qxc5 Re1!, when ...e4 isn't really a threat. Or maybe I'm missing something. It seemed like this was white's best bet, and though I don't know how perfect the master game with Nb3 was white did win the game. I totally missed that.

Nobody likes the game?

I like it.

As a French player myself, I always enjoy to see Black come out ahead! =) I am always thrilled to see White play the Advance variation! Especially if they mishandle it with Bd3 as was the case in this game..  But I usually go for Qb6 right away (instead of Nc6).. Have you looked into that variation?

Thanks padman.

"But I usually go for Qb6 right away (instead of Nc6).. Have you looked into that variation?"

No, probably because my french book never looked at that.

After 6.Bd3 Qb6 is somewhat an inaccuracy. cxd5 first is better since after Qb6 dxc5 is imo more promising than the milner-barry
Move 11 Nb3 is better same with move 12
13...Rae8 is bad. Nxe5 Qxe5 Bxf2! Kh1 (Rxf2 Qxf2 Kxf2 Ng4+ is winning.)Rae8 Qg5 h6 Qg6 Ne4 Bxe4 dxe4 Qxb6 Bxb6 Rxf8 Rxf8 is very grim for White since the e- pawn is going to cost a piece sooner or later
After Rae4 Ndf3 White is atleast equal and maybe even a bit better
After Re5 White
19.Nxe5 Ng3+ 20. Kh2 Nxe5 is according to Rybka -+ (2.30) compared to -+ (1.03) of Nxf1+                            Added the computer variations.

On the first comment, my book seems to disagree (maybe it's just you). Yeah dxc5 didn't really deserve a question mark, because IF it was played correctly (as I mentioned in one of my follow up posts), he can still have decent coverage of the central squares. Still, black should stand well in any case and white certainly let me make full use of my center.

I really appreciate the computer variations. I COMPLETELY missed (after ...Nxe5 Qxe5) ...Bxf2 Qxf2+! followed by ...Ng4.

So once Kh1 is forced, then yes clearly the queen is horribly placed on e5.

Interesting to see that 20...Nxe5 is so strong. Any variations?

I have no idea why I didn't look at Ndf3. Maybe I only looked at the more obvious try f4, which is of course impossible.

I don't own a chess computer and prefer to evaluate by just looking at a position.

Look at the game after 14 moves.  Black has two rooks on adjacent files and
a bishop and a queen pointing at f2. 

Also, white's bishop and rook are out of play.  Further, only 1 out of 4 center
pawns are left and white can't really take advantage of having the only light
bishop.  That position would make a nice chess puzzle!

Black's game is ok after Ndf3, but that blockade on e5 is the thing getting in the way of black's otherwise dangerous initiative. I think I thought that the pin on the e file would be too annoying for white to deal with. Like for example a line like 14 Ndf3 Nxe5 15 Nxe5 Bd6, meeting 16 Bf4 with ...Nh5 maybe, but that would probably be refuted by Nxd7 if the f6 knight moves.