The British election appears to have provided a relatively comprehensive comparison of internet vs telephone polling. It is clear from the data provided in this post from Nate Silver's blog fivethirtyeight.com that traditional telephone surveys still provide a more reliable method of polling than the internet. I think the internet is going to be the way all polls are done some day but it is clear from what happened in the UK that there are unresolved issues with internet polling. The average total error for the telephone surveys was 5.2 and for the internet 11.4. Interestingly, the weakest performance among the major pollsters was the Canadian firm Angus Reid.
Silver's comment: "at this point, the challenges facing Internet pollsters are relatively formidable. Moreover, the bias of the Internet polls -- too many votes assigned to LibDems, who have younger and more-wired voters, and too few assigned to Labour, for whom the opposite is true -- ran in exactly the direction that you might expect."
Canadian firms are moving in this direction. We have had recent internet polls from Decima and Leger in addition to Angus Reid.
Another new methodology starting to gain credibility is IVR, which stands for Interactive Voice Recognition. Ekos in Canada started using this method in 2008 and had the second closest outcome on the 2008 federal election (the online Angus Reid survey was closest but stopped polling several days before the end of the campaign).