|Re: [jetty-users] Jetty PUSH strategy
Hi, On Mon, Dec 30, 2013 at 5:20 PM, Ryan Young <ryanyoung.soloy@xxxxxxxxx> wrote: > I mean at client side we may need to face the case: > > the resource is being pushed, but the data has not yet arrived > > So I wonder how the client deal with this case to avoid sending redundant > "GET" for resource that is being pushed. This is more a question for browser implementers, but I'll give you my version. The client is about to ask for a secondary resource B that may be pushed. It first checks B in its cache. If present and valid, it uses that (and possibly discards the push). Then, it looks at the push cache built for the primary resource. If B is present and valid, it uses that. If B is going to be pushed, but not there yet (your case above), it waits for it. When it is arrived it just renders it on the viewport. Otherwise it issues a GET (which may also contain if-modified-since headers, etc.). There always is a race between what is being pushed and the GET from the client. The server can optimize that and send the pushed headers before or concurrently with the data for the primary resource. Jetty does this, and along with other SPDY improvements, there is a good chance that the reply and the pushed headers are sent in a single write, probably even in the same TCP packet. This reduces the window between the client issuing GETs and the server pushing resources, since the client will know pretty early that pushed data will arrive, and can avoid to GET it. SPDY push is not a fool proof mechanism to push resources; however, in most cases provides a sensible perceived improvement in page "rendering", from our experience. -- Simone Bordet ---- http://cometd.org http://webtide.com http://intalio.com Developer advice, training, services and support from the Jetty & CometD experts. Intalio, the modern way to build business applications.
Back to the top