[Design] Service performance targets

[Jared Rhine]

Brian Moseley wrote:
> webcal GET (modeling ical interop)

In previous performance testing, the times I observed for ATOM GET and
webcal GET were identical within a couple percent.  I assumed that the code
paths were pretty similar (iterate over all events in a collection,
serialize each into a string, and concatenate).  As you understand the
codepaths better, can you see a situation where one of these would slow down
but the other one wouldn't (on a loaded server)?

Thanks for the feedback.

-- Jared