Al Steiner writes in to ask about this position:

Dennis,Bishops of opposite color, 1 extra pawn for black, white to move and….. draw? A few of us at chessbase tried to work this out, and cannot find a win for white OR a draw. Black always seems to get the black bishop and then the win. What do you think?
...If it is truly an odd position, where opposite color bishops don’t draw, then I thought you might be able to use it on “The Chess Mind”. If not, then just chuck it in the recycle bin. Yes, it is my game. I couldn’t find a draw with fritz, and neither could a friend with rybka. Then again, I’m no fritz expert.
We played with it for a while, and it seems the h pawn keeps white’s king on the kingside, and therefore white cannot prevent the outside passed pawn with out giving up his bishop.
Thanks for writing, Al: your letter affords the opportunity to remind the readers of at least three very important points. First, nowhere near all opposite-colored bishop endings are drawn; second, a handy rule of thumb for evaluating whether an ending is a win or a draw is the so-called principle of two weaknesses; third, beware of the glib generalizations of beginners' books. We'll take each point in turn, but first let me offer the bottom line about this position: It's a straightforward win for Black, as the passed h-pawn and pending passed a-pawn stretch White's defenses beyond what they can bear.
(1) Opposite-colored bishop endings
Opposite-colored bishop endings do have a higher drawing ratio than other endings, for two reasons. First, it's often possible for the weaker side to arrange the pawns so they're immune to attack. If the strong side has a light-squared bishop and all the pawns are on dark squares, protected by (a) other pawns, (b) the bishop, and/or (c) the king, then that's the end of that, unless the strong side also has a passed pawn to bother the opponent. Second, when there is a passed pawn, the weaker side can often blockade it on a square whose color is the same as its bishop and opposite that of the opponent's.
(2) The principle of two weaknesses
That's the good news for the defender. The bad news is that there are plenty of exceptions, and the position above is one of them. And here we can combine our first and second main points: not all opposite-colored bishops are drawn, and we can identify that this is likely one of those exceptions on account of White's two "weaknesses".
Let's parse this statement - what does it mean? All of White's pawns are defended and will be safe practically forever, so where are the weaknesses? The answer is that "weakness" is used in an extended sense; here it refers, roughly, to anything a player needs to defend or defend against: a pawn, a square, the king, a queening threat, etc. And White has to defend against two big threats: the h-pawn and the a-pawn. If it was only one or the other, then White could draw in his sleep, but the two together leave him in a hopeless situation. (For some analytical confirmation, click here.)
(3) Rules of thumb
I close with a reminder. Basic books offering rules of thumb are quite useful to beginners, helping reduce the buzzing blooming confusion of the chess board to something more manageable. The sober truth, however, is that the game is very concrete, and exceptions abound. Opposite-colored bishop endings can be won and lost, and in the middle game they often favor the attacking side. It's often fine to move pieces more than once in the opening (think of White's Ng1-f3xd4-b5-a3-c2-e3 in the Sveshnikov Sicilian), to bring the queen out early, to grab "poisoned" pawns, and so on.
These rules of thumb are only tools, not immutable laws. As such, they are our servants, not our masters, and if a tool doesn't help for the job at hand, we should change tools, not jobs!
chessnut: Thanks for the comment about the blog. About the specific claim, it's mysterious to me, for the reasons fluffy mentioned. Honestly, this isn't a difficult ending to work out, generally speaking, once one gets over the beginners' book presumption that opposite-colored bishop endings are drawn in all but rare cases.
Alex: I'm not sure what you're talking about, but you might want to work through my analysis before working out these problem-like possibilities. The only complexity in the position arises if White is given the chance to achieve c4 without losing a pawn first.
Perseus: There's no reason to resign immediately - unless you're playing a strong player, your opponent might misplay it with Black, which is just what happened in the game (45.Bd4?). The comment about removing the a- or h-pawn is correct but probably obvious to all; a more interesting question for those of you trying to work out everything that's happening here is what happens if the f-pawns are eliminated, or if they are pushed up to f6 and f7.