Le jeudi 16 janvier 2014 18:31:59 UTC+1, Boris Zbarsky a écrit : > It's pretty far away in Gecko as it stands, because we have all sorts of > global state (class statics, say!) that is not parallel-safe. > > Process-separated is likely to be simpler to get to work without pesky > things like memory corruption. I don't understand the last sentence. Is "memory corruption" referring to the global state you mentioned before? When you wrote "parallel-safe" above, was that implicitly thread-parallelism?
As a web developer, I'm happy with process-separated if it's easier, at least for a first shot, but I can imagine the costs aren't the same at scale. Maybe the current rarity of sandboxed iframe on the web makes it a practical idea? I forgot to mention that Blink is planning [1] to work on site isolation [2] (a site is an eTLD IIUC). I think it flows that sandboxed iframes will be process-isolated (since a sandboxed iframe is in its own "site"). I believe this lowers the chance of Blink adopting the WebWorker-specific APIs to draw on canvas/WebGL contexts that Mozilla is working on. David [1] https://groups.google.com/a/chromium.org/d/msg/blink-dev/Z5OzwYh3Wfk/IWooaY5FZowJ [2] http://www.chromium.org/developers/design-documents/oop-iframes _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
