We have several features that interact with some cache that is not really defined: rel=prefetch rel=preload H2 server push I think the approximate semantics we want are HTTP cache semantics. However, resources that cannot be cached shoul...
It then uses a greedy algorithm to identify the best time points to use given a limited budget (i.e., if the user can only profile x number of time points). TPS can also estimate the resulting error from using less time points and so provides a way to balance accuracy and costs. ...