1. The Liberals did exceptionally well in Toronto, poorly out west. Can one generalize about Dion on the numbers? Doesn't look like it. TC's guess is local factors were important. The candidate controversy in Saskatchewan hurt the Liberal. Turnout differential within the riding may well have helped the Conservative. The Liberal candidate in BC seemed to be weak.
2. Bad night generally for the NDP. Their vote held up better in the Saskatchewan and BC than Ontario. Their effort appeared to be minimal.
3. Good night for the Greens. It lends credibility to polls showing them doing well. They look like a continuing problem for both the Liberals and the NDP. However, the close outcome in Quadra will likely lead to strategic voting in the general election that boosts Liberal numbers next time. The Greens continue to have the least efficient vote. In most of TC's forecasts in the past few years they don't win a single seat. Their leader's claims (that they can win 6 to 12 seats) should be discounted.
4. Mixed night for Harper but winning even one seat is a boost. Loss in Quadra, which could have gone the other way, is a blow. His inept Finance Minister probably cost them votes in Ontario.
5. TC has a method to project province-wide numbers from the individual riding results. Here is a table that displays them and calculates province-wide seat numbers from the projection (need to be taken with a grain of salt).
UPDATE: Well, not enough. The by-election numbers are essentially consistent with this latest Harris-Decima survey.