I agree that Blink-in-JS and V8 extras have different goals. V8 extras is supposed to enable exactly what you describe without impacting the performance of the feature in a negative way (e.g., like a V8 extension that would incur additional costs for each new context due to compilation and execution).
You know that I disagree about Blink-in-JS's utility for implementing web platform features, so I won't repeat that here in detail again. Design doc sounds indeed like a good idea. Domenic, you had an internal doc about streams + v8 extras. Mind brushing that up and making it public? thanks -jochen On Tue, May 19, 2015 at 9:53 AM Kentaro Hara <hara...@chromium.org> wrote: > This sounds pretty great, but I'm also curious about the relationship >> to Blink-in-JS. > > > I'm curious about this as well -- in more general, I'm curious about how > we're going to layer the web platform. > > Here's my understanding; correct me if I'm wrong. > > - V8 extras and Blink-in-JS are aiming at different goals. > > - Blink-in-JS is aiming at implementing WebIDL features on top of existing > Web APIs. Thus we don't want to add any "backdoor"s to the Web APIs, since > those backdoors will break the layering philosophy of Blink-in-JS and make > it harder to port the implementation to other vendors. > > - V8 extras is not directly related to WebIDL and is a more intrusive way > to implement any feature using JS (instead of C++). For example, you can > implement streams etc in V8 extras. If the streams.js needs to interact > with C++, you can just expose V8 APIs for the interactions and the binding > layer to connect the V8 APIs with the C++ (e.g., V8Streams class that > abstracts the V8 APIs etc). From the viewpoint of Blink, the feature > implemented by V8 extras will just look like other features in V8 -- in > other words, there are V8 APIs to interact with the feature and Blink can > just use the V8 APIs through the binding layer. This is a layering > technology inside Chromium and it is not planned to port the JS > implementation to other vendors. > > > (I also want to see a design document of V8 extras.) > > > > On Tue, May 12, 2015 at 8:43 PM, Philip Jägenstedt <phil...@opera.com> > wrote: > >> This sounds pretty great, but I'm also curious about the relationship >> to Blink-in-JS. >> >> I poke at our <video controls> implementation in C++ from time to time >> and it's just screaming out to be rewritten in JavaScript, so I'd like >> to know if V8 extras might be useful for that. What's needed is a way >> for C++ to hand a video element and an empty container to external >> JavaScript code that would add controls to the container and hook up >> the events. >> >> On Tue, May 12, 2015 at 2:20 AM, Kentaro Hara <hara...@chromium.org> >> wrote: >> > Is the V8 extras going to be used for features that can completely >> > implemented on top of JavaScript and already existing DOM APIs? Or do >> you >> > have a plan to make the V8 extras interact with C++ DOM somehow? (If >> so, I'm >> > curious how you're going to implement the binding layer.) >> > >> > >> > >> > On Tue, May 12, 2015 at 3:52 AM, Domenic Denicola <d...@domenic.me> wrote: >> >> >> >> Over the last two weeks we’ve been working on a new feature designed >> for >> >> V8 embedders, called “V8 extras.” The intent is to allow embedders to >> write >> >> self-hosted (i.e. JavaScript) implementations of various features, >> with more >> >> power and less drawbacks than the existing V8 extensions mechanism. >> Notably, >> >> we are building on top of recent snapshotting work to ensure that the >> code >> >> for V8 extras is included in the snapshot, and so should not have >> negative >> >> impact on context startup time. >> >> >> >> You can see a working, albeit in-progress, example of a V8 extra in >> this >> >> CL, which uses them to implement readable streams from the Streams >> Standard. >> >> General features on display include: >> >> >> >> The inclusion of the extras via the v8_extra_library_files GYP variable >> >> The use of V8 native functions, such as %CreatePrivateOwnSymbol, in the >> >> .js files (but see below) >> >> The use of the extrasExports object to export functions and values to >> the >> >> embedder. This object can be accessed through the V8 C++ API with >> >> v8::Context::GetExtrasExportsObject(). >> >> >> >> There are a number of changes we’re still working on implementing >> before >> >> we could comfortably say that this feature is ready: >> >> >> >> We do not want to allow extras direct access to V8 native functions, >> but >> >> instead only to a small number of pre-specified ones. The details on >> how to >> >> do this are still being worked out, but it will probably involve a >> slight >> >> change to the authoring format of the extras files, to become e.g. >> script >> >> whose completion value is a function that gets such a utility object >> passed >> >> in. >> >> We need a story for segregating “experimental” extras from shipping >> ones, >> >> similar to what is being done in V8 already for harmony features. >> >> Experimental extras cannot live in the snapshot, since they must be >> able to >> >> be turned off at runtime, while shipping ones will be compiled in. >> >> We eventually want to expose an easy way for the embedder to supply >> >> functions or values to the extras code. This is not actually needed for >> >> streams, but for example if we were to implement something like >> TextEncoder >> >> or URL as an extra, it would be crucial. >> >> >> >> Overall though, we are quite excited about the possibilities this opens >> >> up, by allowing embedders a performant way to implement features in >> >> JavaScript. >> >> >> >> Best, >> >> >> >> Domenic Denicola >> >> Jochen Eisinger >> >> Yang Guo >> > >> > >> > >> > >> > -- >> > Kentaro Hara, Tokyo, Japan >> > >> > To unsubscribe from this group and stop receiving emails from it, send >> an >> > email to blink-dev+unsubscr...@chromium.org. >> > > > > -- > Kentaro Hara, Tokyo, Japan > > -- > -- > v8-users mailing list > v8-users@googlegroups.com > http://groups.google.com/group/v8-users > --- > You received this message because you are subscribed to the Google Groups > "v8-users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to v8-users+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout. > -- -- v8-users mailing list v8-users@googlegroups.com http://groups.google.com/group/v8-users --- You received this message because you are subscribed to the Google Groups "v8-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to v8-users+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
