When setting transitionTimingFunction as "linear":
Safari 5.0.1 Mac returns "cubic-bezier(0, 0, 1, 1)"
Chrome 8.0.552.23 dev Mac returns "linear"
Tests were written to expect "cubic-bezier(0, 0, 1, 1)", so they fail in Chrome
Firefox 4.0 beta Mac returns "border 1s cubic-bezier(0.000000, 0.000000, 1.000000, 1.000000)"
is it reasonable to expect "linear" back if you set it as "linear"?
I suggest normalizing on the cubic-bezier() functions. All the named shortcuts come with official definitions in the spec so we can easily map name --> function params. From there if we normalize the params by ramming them through Number() we will probably wind up with predictable values.
The one possible danger I see in doing so is possible floating point oddness. I'd be surprised frankly to see it come up, but it's something to watch out for.