Hi Am 25.03.2015 um 20:28 schrieb Dave Huseby: > I wonder if we created daemons that wrapped all of the vendor binaries > if we could get a situation where vendor FOTA's only included their > binaries and customized versions of our daemons and did NOT include > gecko? If we could get there, then the vendors should not have any > problem with us updating gecko independent of their update system.
That sounds great, but I'd guess that phone vendors don't want to ship code that they haven't tested extensively. And testing takes time and costs money... I could also imagine that phone vendors continue to provide updates as they do now (or sometimes do not). And Mozilla or our community provide updates to the latest versions of FirefoxOS. Users would be free to choose either (stable vs latest-and-greatest). The overall update process can certainly be automated to a good extend, so we'd serve a large number of different phones with a small overhead. > As long as there was a robust versioning system for the interface > between our wrapper daemons and gecko, then we'd be able to detect > compatibility between what's on the phone and the version of gecko we > want to update to. Let's assume we have a RIL daemon that can wrap a vendor's binary RIL component. For RIL we use the IPC and protocol code that comes with Android. I don't know about feature detection, but Gecko's RIL code is currently compatible with multiple versions of Android. I don't think we'd run into compatibility problems here as the RIL daemon could probably implement any version. And for the RIL daemon itself, we can detect or configure the vendor's version during B2G's build process. The build scripts would fetch libxul.so and vendor binaries (e.g., RIL XPCOM components) from an existing phone, detect their version or feature set, and build the daemon against them. IANAL, but I think that as long as we don't redistribute the vendor's binary code, this should also be OK from a legal perspective. I really don't know if any of what I wrote is possible in practice. I'd guess that wrapping a vendor binary component behind an IPC protocol is not as simple as it sounds. Best regards Thomas > > --dave _______________________________________________ dev-b2g mailing list [email protected] https://lists.mozilla.org/listinfo/dev-b2g
