Hi Sean On Fri, May 12, 2017 at 10:33 PM, Sean Whitton <spwhit...@spwhitton.name> wrote: > > Dear Sebastian, > > On Thu, May 11, 2017 at 02:29:35PM +0200, Sebastian Noack wrote: > > as you might know, Firefox is moving away from traditional Gecko > > extensions, towards WebExtensions which are essentially the same as > > Chrome extensions. > > Right. We need to update our tools, but the work hasn't started yet.[1]
As far as I understand, the only thing that xpi-pack does, besides creating an XPI file (which essentially is just a standard ZIP file), is some magic for unpacked JAR files within, which is no longer relevant for WebExtensions. Moreover, creating an XPI/ZIP file, just to have it unpacked by install-xpi seems redundant in the first place. Besides that, the only thing that install-xpi seems to do is creating symlinks in /usr/share/mozilla/extensions/ based on the supported applications listed in install.rdf. But WebExtensions should not have an install.rdf, and support exactly one Gecko application (i.e. Firefox). So while xpi-pack and install-xpi seem useless for WebExtensions, what might still make sense are the ${xpi:*} substitutions in the control file, but the minimum/maximum Gecko version would have to be extracted from manifest.json instead of install.rdf. > > WebExtensions don't need an install.rdf manifest. In fact the > > documentation indicates that is has been deprecated as part of legacy > > extensions [1]. Still, I would have to create an install.rdf, just so > > that I can use install-xpi. > > > > However, it is much simpler, and probably not any less appropriate > > than relying on deprecated mechanisms, to simple use debian/install > > and debian/links files, in order to achieve the same result, as I did > > here [2]. > > > > But I wonder whether this is the way how Firefox extensions are > > supposed to be packaged in the future. Otherwise, mozilla-devscripts > > would have to be updated, in order to support WebExtensions properly, > > unless I miss something. > > > > Also I wonder whether it still makes sense to package Firefox and > > Chromium extensions separately, if the the only difference is either a > > symlink in /usr/share/mozilla/extensions/ or a script with one line in > > /etc/chromium.d. > > Thank you for these suggestions. It would be great if Firefox and > Chrome addons could just be under a webextension-* namespace. FWIW, I'd rather go for the abbreviation "webext-" as this consistent, both, with the old "xul-ext-" prefix, and with the "webext" W3C group which standardizes these. But this is just bikeshedding, I suppose. Anyway, this generally seems to be a good idea, though there are some more things to think about: * Where should cross-browser extensions be installed in the filesystem? * What is about extensions that will remain only compatible with either Chromium or Firefox? * What is about extensions that support both browsers, but with different builds? > I think the best thing might be to have a meeting about this with any > pkg-mozext people at Deb[Camp|Conf], to figure out a transition plan. That sounds like a good idea and/or perhaps we could loop somebody of them in on the discussion here. Sebastian
_______________________________________________ Pkg-mozext-maintainers mailing list Pkg-mozext-maintainers@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/pkg-mozext-maintainers