OK, time to make the rubble bounce. There's a nice online visualiser for climate data here, (thanks!) so we can have a lovely little chart.
OK, this shows the GISTEMP land-ocean mean for as long as they've been measuring, as the red line. As you can see, that sucker's going up. As you can also see, the variance is considerable - it bounces about quite a lot. The green line shows the trend. Over on the left, I've plotted trendlines starting 1998, dark blue, and 1999, purple.
As you can see, the hot year 1998 makes the trend look flatter over the next ten years - but it's still upwards, because this is a trend estimate, not just a line drawn from one end of the plot to the other. The computer wouldn't do anything that dishonest. But just to illustrate it, I've added a trend plot from 1992 to 2006, in light blue. Scary, huh? Going up like a rocket.
Of course it is; because it's completely meaningless. I selected those dates because 1992 was an unusual cold year caused by a volcanic eruption and 2006 was hot, which is no different to picking 1998 as a start point because it was hot.
If you do what Duff did, and forget that it is now 2009, as you can see from the chart, the trend in the last ten years would be going up FASTER than the trend across the whole dataset.
Someone arguing in good faith would have immediately dropped their sublime confidence at 0001Z 01/01/09 and started buying inland property, bullets, and toilet paper, to say nothing of apologising to the world at large.