Now, it's better to spend those 3 years iterating than planning - rumor has it that the GMail prototype was built in a day. But it's worth being realistic about how long it takes to build a world-changing product - it's a long road. (In fairness to Marissa, I'm sure she knows this, and I suspect she's just trying to get the release cycle down so that Yahoo can make forward progress.)
Apparently it was available internally http://en.wikipedia.org/wiki/History_of_Gmail
It certainly does not take 3 years to produce GMail, if they had the focus.
Maybe this is not the right time for Yahoo to do 3 year products simply because they suck at it (I spent more than 3 years at Yahoo, and I spent most of that time trying to shepherd projects through to the point my engineering team would actually get a go-ahead; we were a service function without direct control of a product, otherwise we should've just gone for it rather than wait...), and practising on 6 month projects could still bring a ton of useful improvements or smaller services. One thing to keep in mind with Yahoo is that it has a vast array of sites and services just sitting there - I bet there's a ton of poorly monetized services that could get drastic upgrades in well below 6 months if there's just sufficient fire behind someones asses...
Then they can start trying to add the standalone world-changing stuff again later, if they're successful.
Looking from the outside, they seem to have a dysfunctional culture around shipping software with all the additional layers of management. Hopefully this changes and they have a chance producing meaningful products but I'm not hopeful.