Conventional Caro-Kann theory holds that Black should trade here, as this eliminates White's good bishop, whereas Black's light-squared bishop will be less valuable after he seals the centre with e6.
6. Qd1xd3 e6 7. Ng1f3 Nb8c6
Black must keep control over e5 and c5, hence the queen's knight move here.
This puts White's bishop on a strong diagonal, controlling the crucial e5 square.
8... Ng8f6 9. h3
This looks like an unnecessary precaution, as Black had nothing to gain by posting a knight to g4 anyway.
Conventional C-K theory holds that Black must dispute the power of White's bishop by playing to d6 here. I wished to hang on to the good bishop for a while, and therefore chose to play the dark bishop to e7 instead.
This looks like an attack now. Maybe h3 last move was not precautionary after all.
Black decides to castle into it and batten down the hatches. When defending against a pawn-slaught, the defender should avoid pawn moves of his own. This forces the attacker to expend several more tempi to get his own pawns into exchanging range to open up lines of attack for the heavy pieces; whereas any pawn move by the defender simply makes it easier for the attacker to pick them off and open up lines of attack.
This prevents the Black knight from coming to b4 and driving the White queen off the powerful b1-h7 diagonal.
This is a fairly typical Caro-Kann sort of position: White aims to control the crucial squares e5 and c5, whereas Black aims to post pieces to the crucial squares e4 or c4. The rook move to c8 prepares to control c4.