- From: Ilya Grigorik <igrigorik@google.com>
- Date: Fri, 23 Jan 2015 09:33:24 -0800
- To: "Schurman, Eric" <ericsc@amazon.com>
- Cc: "public-web-perf@w3.org" <public-web-perf@w3.org>
- Message-ID: <CADXXVKqQbDNNo_yxcaj-dbj3Mxowyj5ury7cgqj3yW7fh8ED0A@mail.gmail.com>
On Thu, Jan 22, 2015 at 3:00 PM, Schurman, Eric <ericsc@amazon.com> wrote: > Has there been any more thought on deprioritizing assets? > Being able to apply to specific elements something like the "lazyload" > attribute specified in Resource Priorities ( > https://dvcs.w3.org/hg/webperf/raw-file/tip/specs/ResourcePriorities/Overview.html) > would have directly met some needs that we have. Subtrees would be great > too. > "deprioritization" has many facets: network priorities and dependencies, deferred fetching, execution behaviors, etc. Given the number of various permutations and requirements, I think we should be focusing on enabling the smallest number of low-level primitives that enable applications to implement (and experiment with) these behaviors on their own. In terms of required pieces: 1) Provide API's for setting/managing fetch-layer settings: e.g. request priority, dependency trees (in http/2), etc. 2) Separate fetching from processing to allow applications to define own behaviors. For (1) we have several threads on whatwg about exposing new APIs on fetch() and providing a similar mechanism via 'fetch-settings' attribute on declared elements. For (2), I believe rel=preload is the answer, as it separates download from processing and yields full control to the application for how/when fetched resource is applied. In short, I think we have the necessary pieces in the pipe... ig
Received on Friday, 23 January 2015 17:34:31 UTC