On 02/26/2015 10:58 AM, Brian Leathem wrote:
On 2015-02-26 07:13 AM, Summers Pittman wrote:
> I really, really, REALLY wish Google hadn't made Android a class and
> had gone for an injection/composition programming model instead.
> It's like they weren't even paying any attention to academic and
> professional literature about OO and Java in 2006...
Agreed - it's a severe limitation indeed.
In terms of landing the google-play-services oauth2-token request
feature I am trying to implement, I see a number of paths forward:
1) Rewrite the Aeroegear Oauth2 API altogether to be intenet based,
providing users the token via the #onActivityResult method. This
would mostly just be adding a layer in front of the existing API, and
shouldn't be too honerous. It would however be a breaking change for
downstream applications.
2) Leave the google-play-services request out of android-authz
(Android users can initiate the play-services intent themselves if
they want). Instead I would implement this feature only in the
oauth2-codova plugin where I can receive the intent response.
3) Explore using an Android Service to handle Oauth2 token requests.
I believe we would be able to trigger an intent and retrieve the
result from within the service, then provide it to the downstream
application from the service. It's a little unclear as to whether
starting intents from a Service is good practice.
My vote would be to pursue 2) for now, with a note about 1) being
added to AGDROID-319 to be treated in a future revision.
I agree. For 1 we should
leave the old implementation as an option.
Not having your activity die during log in is a good feature and the
dialog is the easiest way I can think of it and it keeps it from being a
breaking change.
Thoughts?
--
Summers Pittman
>Phone:404 941 4698
>Java is my crack.