incubator-callback-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian LeRoux...@brian.io>
Subject Re: endless refactoring of plugins until "Cordova 2.x"
Date Mon, 26 Mar 2012 18:43:48 GMT
I'm far more in favor of shims that barf deprecation notices than
maintaining two branches.

Bryce/Patrick: I'll kick up a separate thread about roadmapping plugin stuff.


On Mon, Mar 26, 2012 at 11:31 AM, Shazron <shazron@gmail.com> wrote:
> I haven't done any JavaScript shims but I've eased the pain for users
> needing to use plugins by adding compatibility headers for all the
> PhoneGap/Cordova versions out there (0.96 to 1.5.0). See my plugin in
> this issue and its README:
> https://issues.apache.org/jira/browse/CB-330
>
> These compat. headers I include in CordovaLib but is so far
> undocumented. What the procedure is, you would write a plugin meant
> for 1.6.0 (ie the latest version), and for older version support, a
> dev would included the appropriate compatibility headers.
>
> It's working great so far for the CDVLocalStorage plugin - from the
> same plugin code base (one .m and .h) through the compat. headers,
> everything just works with older versions as well. I assume we can do
> something similar for JavaScript as well (shims).
>
> Shaz
>
> On Mon, Mar 26, 2012 at 11:15 AM, Bryce Curtis <curtis.bryce@gmail.com> wrote:
>> I agree that there's no way to know what plugins will look like in 2.x.
>>  Until then, I strongly believe we need to maintain the same (undocumented)
>> API that plugins currently rely upon, with the goal that 3rd party plugins
>> won't need to be updated again until 2.x - They already took a hit with
>> renaming in 1.5.
>>
>> Regarding the plugin APIs (JS & native), we can't identify with certainty
>> which ones will be deprecated until we are further along with the 2.x
>> plugin architecture, so we shouldn't confuse users until details are worked
>> out and prototypes running.  Regarding API stability, while we have
>> discussed deprecating methods like addConstructor, window.plugins, etc.,
>>  the fact that they have remained consistent for many releases (well over a
>> year or 2) has led to their wide use.
>>
>> After 1.6.0 release, we should schedule a plugin meeting to go over the
>> priorities, what's been done to date, and where it's going.
>>
>> On Mon, Mar 26, 2012 at 1:05 PM, Filip Maj <fil@adobe.com> wrote:
>>
>>> Thanks Brian, I was just going to respond saying the same thing but you
>>> beat me to it!
>>>
>>> >What I'm trying to say is that I'm not at all inclined to +1/-1 until
>>> >we have agreed on the particulars of the change we seek.
>>>
>>> ^^ The above is the underlying issue. I suggest we do that as a first step.
>>>
>>> The referenced bug in this thread is just a small part of the overall
>>> work. Side note: deprecating + removing the "PhoneGap" JS global is not
>>> necessarily for plugins, it's a part of the Apache rename. The only
>>> changing part of the bug is removal of window.plugins. I prototyped a
>>> deprecation approach in a branch here (diff view):
>>> https://github.com/filmaj/incubator-cordova-js/compare/masterŠdeprecation
>>>
>>> Re: addResource/hasResource. These do not exist in cordova-js right now.
>>> They are effectively gone. They were only there for BB + Android anyways.
>>> I had one dude on IRC complain that we removed that in 1.5.0 without first
>>> deprecating. No huge outcry. Not saying that the fact we just axed it is
>>> good - it wasn't. We should have deprecated first, no doubt. My point
>>> about these two specific methods is, it already happened last release, so
>>> it is a moot point to discuss.
>>>
>>> Additionally we have had some work done to try to prototype what a final
>>> plugins-based approach to app development would look like, specifically
>>> Andrew Lunny's "pluginstall" work (which I believe started from a
>>> thread/discussion that happened on the list between Andrew and Pat). Here:
>>> https://github.com/alunny/pluginstall
>>>
>>> Andrew is most of the way there in terms of defining an XML file
>>> describing the integration of native source for a plugin, see:
>>> https://github.com/alunny/pluginstall/blob/master/test/plugin/plugin.xml
>>>
>>> The native plugin architecture has been around for a while. It is stable.
>>> Not much needs to be done there.
>>>
>>> The biggest question in my mind is how we want to handle the JavaScript.
>>> With cordova-js already implementing a basic module system with a hard API
>>> for the exec() function, we're actually not too far away.
>>>
>>> We should keep going in enumerating all these things that make up the
>>> plugin goal in this thread and drop whatever else comes up into a wiki
>>> page.
>>>
>>> It seems like it is a "safer" idea to slate all the plugin changes / API
>>> removals for 2.0. That being said, can we agree to drop deprecation
>>> notices into the agreed-upon APIs that will be axed leading up to 2.0?
>>>
>>> > I think we
>>> >can agree on the spirit of the focus of the work being a world of
>>> >plugins and tooling for automation. We haven't added much outside of
>>> >battery (and a new platform).
>>> >
>>> >1. The plugin architecture remains completely undocumented.
>>> >2. We do not support 3rd party plugins.
>>> >3. There is no automation or tooling.
>>> >
>>> >Are plugins from an API standpoint stable today?  (I'm guessing not
>>> >when I think of things like addConstructor.)
>>> >
>>> >If they aren't stable, undocumented, unsupported by our effort, and a
>>> >work in progress for tooling: why are we concerned with breaking them?
>>> >
>>> >(Take all above with grain of salt, I think having a 2.x branch a good
>>> >idea, but will slow us down for no direct benefit to Cordova that I
>>> >can currently see.)
>>>
>>>

Mime
View raw message