So... my basic view (and its the correct one, ha ha) is that sea ice is an insulator, not a heat sink. In the winter ice-covered arctic, sea ice sits on top of an ocean whose temperature is about -1.8 oC, whereas the air overhead is about -30 oC (or something, I haven't bothered to check. Cold, anyway). If you cleared the ice away, the ocean would freeze over, of course. Heat conduction through the ice is slow and has more-or-less stopped by the time the ice has got to 1-2m thick (which is why ice thickness doesn't keep going up: the thicker ice comes from ridging). Hence, it acts as an insulator. Insofar as the teleological terms make sense, this is its climatic "function".
Lubos uses the old "google trick" to prove that people think the sea ice is a heat sink: just google "heat sink" "sea ice" he says. But this is a very old trick, and, as ever, throws up piles of hits with "sea ice" and with "heat sink" but no great relation between them. Of the hits, the top one is just the Indie article we're talking about. #'s 2,3 are quoting "The Arctic is the heat sink of the Northern Hemisphere" and *this* is fair enough (because of geometry, the tropics get more heat in, so heat flows from tropics to poles, so the poles are sort-of a heat sink). But thats the whole arctic, not the sea ice.
To point out the obvious, that Lubos seems to have missed: GCMs of course include the latent heat of sea ice in their calculations.
In fact, much of this disagreement rests on arguing about different things. If you are modelling sea ice, then yes of course you need to include the latent heat (if you don't, there is no control at all on whether the ice forms or not). And yes, the latent heat of fusion of a meter or so of ice is significant. But...
Trying to seperate out in more detail the heat capacity/latent heat stuff is more difficult. If you regard the ice as there, then the heat capacity is negligible (looking up the numbers, it seems to be about half that of the same mass/volume of water). If the ice isn't there, then yes its formation will indeed soak up heat. But... taking these fluxes in isolation won't work.
If you want to *store* heat then you clearly can't do that in the ice (since it would melt). And you have to take into account the seasonal cycle. To store it, it has to go into the oceans. And heat storage in the oceans is why the worlds warming is least in the great Southern Ocean around Antarctica and predicted to stay that way for a while.
Summary: calling the sea ice a heat sink, as the Indie does, appears to be a mis-paraphrase of "the *arctic* is a heat sink, which is what people do indeed say. Greg Kuperberg (who seems to do his best over at Lubos's comments) has already pointed this out.
[Update (sorry PF): Brian J points to a nice table from a recent Levitus paper (sorry JA). Then (with large caveats, but emphasising that they are probably overestimating the fusion term) they get for the change in heat storage) 18e22 for the ocean; 6e21 for the atmos (1955-96); misc other terms; 5e19 for the arctic sea ice (1978-96).]
[Update #2: I've struck out the 30m calc. I'm almost sure its out by a factor of 1000... somewhat surprised no-one has ripped it apart yet... I'll check in the morning]
[Update #3: For those of you visiting from Lubos Motls blog, let me point out that his version is seriously misleading and dishonest. The erroneous sentence which he bases his entire post around was struck out before he ever read this; I think he is trying to give the impression that I corrected my post based on his reply; but no. Also, Arun (see comments on this post, about #7) provides a nice link+quotes on the climate role of sea ice.
Meanwhile, for those interested in the status of string theory, this is interesting.]