8:23 a.m.
ok, Kris had some thoughts on a better flow, so i refactored the code a bit and i think i
like this way a bit better.
New Flow - Client Flow - Standalone for now, possible integration with pipes
First Time - No Access Token stored( in localStorage )
User will create the Authorization Object stuff with settings/options
var thing = AeroGear.Authorization();
thing.add({
name: "coolThing",
settings: {
clientId: "12345.apps.googleusercontent.com",
redirectURL: "http://localhost:8000/redirector.html",
tokenValidationEndpoint:
"https://www.googleapis.com/oauth2/v1/tokeninfo",
authEndpoint: "https://accounts.google.com/o/oauth2/auth",
revokeURL: "https://accounts.google.com/o/oauth2/revoke",
scopes: "https://www.googleapis.com/auth/userinfo.profile",
prompt: "force"
}
});
should have the ability to specify more settings, based on the spec
The user would then call some method( currently not good names are coming to me, maybe
validate ) that takes success and error callbacks.
thing.services.coolThing.validate({
success: function( response ){
console.log( "Should be response from Validating the access token",
response );
},
error: function( error ) {
//should contain a constructed URL for the user
console.log( "error", error );
}
});
Since this is the first time, the error callback will be called and will contain the
constructed URL that the user should do the popup redirect dance with to get an access
token.
what "dance" they do is up to the developer
Once that happens and they have the access token, they would call the validate method
again.
this makes sure that the token they recieved is validated and will also return some other
meta data related to the token, like refresh time.
Once the token has been validated, it will be stored in localStorage and would be
accessable with the key of ag-oauth2-whatever_the_client_ID_is .
so in this example it would be something like:
ag-oauth2-12345.apps.googleusercontent.com
There is one problem i can see here though. If the user has to applications with the same
client ID but different scopes assigned, this would be a problem. That use case could be
considered bad practice anyway
The user can then call the "callService"( yes, again, crappy name ) method to
get access to the service they want.
thing.services.coolThing.callService({
serviceURL: "https://www.googleapis.com/oauth2/v2/userinfo",
success: function( response ){
console.log( "Should be the response from the call", response );
},
error: function( error ) {
console.log( "error", error );
}
});
All these methods would have success/error callbacks.
Token Expiration
If the user makes a call to a service, using the callService method, and they recieve an
error such as not authorized or token invalid or token expired, I'm thinking we send
what the "contructed URL" should be, similar to the validate method described
above.
Since this is a Client Side flow, there is no refresh token, so the client wouldn't be
able to refresh the access token without doing the "dance" again.
On Aug 27, 2013, at 1:57 PM, Lucas Holmquist <lholmqui(a)redhat.com> wrote:
i've hacked together a sample app that shows sort of the flow.
https://github.com/lholmquist/oauth2test
it is still very rough
On Aug 27, 2013, at 12:42 PM, Bruno Oliveira <bruno(a)abstractj.org> wrote:
> +1 keep it simple, please
>
> Lucas Holmquist wrote:
>>
>> On Aug 27, 2013, at 3:39 AM, Sebastien Blanc <scm.blanc(a)gmail.com
>> <mailto:scm.blanc@gmail.com>> wrote:
>>
>>> Hi,
>>> That sounds good !
>>> Just one question, instead of using the callApi function couldn't we
>>> pass the oauth module (called 'thing' in your example) to the pipe
>>> directly, using the 'authenticator' setting. Behind the scene, the
>>> pipe manager will append the oauth token to the query or add the
>>> bearer header ?
>>
>> I'm not sure if that is what this is going to do. This is more of an
>> Authorization thing and i don't think it totally fits the pipeline
>> stuff. ( or it would make it a bit more complicated, and we want to keep
>> it simple )
>>
>>
>> i should probably change the method to be "authorize" instead
>>
>>> Seb
>>>
>>>
>>>
>>> On Mon, Aug 26, 2013 at 8:05 PM, Lucas Holmquist <lholmqui(a)redhat.com
>>> <mailto:lholmqui@redhat.com>> wrote:
>>>
>>>
>>> OAuth2 AeroGear Workflow - High Level
>>>
>>>
>>> Using Google api's
>>>
>>> /Server Side/
>>>
>>> 1. user needs to first create an "application/project" to get
an
>>> api key
>>> 2. Then they would choose the services/api's then would like
>>> there application to access
>>> 3. other google server related items....
>>>
>>> /Client Side/
>>>
>>> 1. Create a new OAuth2 module thing
>>> 2. Get access token for the services would need to specify the
>>> services they would like to access
>>> 3. validate the token
>>> 4. make calls to the service
>>>
>>>
>>> API
>>>
>>> |var thing = AerGear.OAuth2({
>>> name: googleEndPoints, //Just a Name
>>> clientID: "12345" //The client ID of the app
from the API console
>>> settings: {
>>> permissions: "..",
>>> ...
>>> }
>>> }).somecoolmodulename.googleEndPoints;
>>> |
>>>
>>> /Settings: Multiple settings based on paramters here
>>> <https://developers.google.com/accounts/docs/OAuth2UserAgent>/
>>>
>>> /Methods/
>>>
>>>
>>> authenticate
>>>
>>> this will authenticate with the server to get the access token and
>>> then validate the token, once that is all good then the response
>>> is returned.
>>>
>>> |thing.authenticate({
>>> success:{},
>>> error:{},
>>> settings: {
>>> //probably some settings here, like URL overides and such
>>> }
>>> });
>>> |
>>>
>>>
>>> callApi
>>>
>>> not really a good name, but it would basically call the remote
>>> api/services. we could either do a query string option or a Head
>>> option
>>>
>>> example:
>>>
>>> |curl
'https://www.googleapis.com/oauth2/v1/userinfo?access_token=1/fFBGRNJru1FQd44AzqT3Zg'
>>> |
>>>
>>> or
>>>
>>> |curl -H "Authorization: Bearer {accessToken}"
https://www.googleapis.com/oauth2/v1/userinfo
>>> |
>>>
>>> code:
>>>
>>> |thing.callApi({
>>> service: "userinfo", //don't really like this name
either
>>> success:{},
>>> error:{},
>>> settings: {
>>> ... //overridable baseURLs?
>>> }
>>> });
>>> |
>>>
>>>
>>> revoke
>>>
>>> again, maybe not the best name. calls the "revoke" service, to
>>> remove access to permissions
>>>
>>> |thing.revoke({
>>> success: {},
>>> error: {},
>>> settings: {}
>>> });
>>> |
>>>
>>> Behind the scenes on all these calls, the "access_token" is
>>> beining used and possibly refreshed for the user, so they don't
>>> have to worry about it. They just need to call authenticate first.
>>> Maybe we can have a refresh method if the user wants to refresh
>>> the tokens themselves. this would do the token "dance"
>>>
>>>
>>>
>>> On Aug 26, 2013, at 1:35 PM, Bruno Oliveira <bruno(a)abstractj.org
>>> <mailto:bruno@abstractj.org>> wrote:
>>>
>>>> +1 I think is a good start to us.
>>>>
>>>> Kris Borchers wrote:
>>>>> I would like to see that but what you are saying makes sense. It
>>>>> sounds like where I was headed with the Basic and Digest
>>>>> adapters before I ran into browser security issues with headers.
>>>>> I think and authorization API that basically just wraps itself
>>>>> around secured endpoints works for me.
>>>>
>>>> --
>>>> abstractj
>>>>
>>>>
>>>> _______________________________________________
>>>> aerogear-dev mailing list
>>>> aerogear-dev(a)lists.jboss.org
<mailto:aerogear-dev@lists.jboss.org>
>>>> https://lists.jboss.org/mailman/listinfo/aerogear-dev
>>>
>>>
>>> _______________________________________________
>>> aerogear-dev mailing list
>>> aerogear-dev(a)lists.jboss.org <mailto:aerogear-dev@lists.jboss.org>
>>> https://lists.jboss.org/mailman/listinfo/aerogear-dev
>>>
>>>
>>> _______________________________________________
>>> aerogear-dev mailing list
>>> aerogear-dev(a)lists.jboss.org <mailto:aerogear-dev@lists.jboss.org>
>>> https://lists.jboss.org/mailman/listinfo/aerogear-dev
>>
>> _______________________________________________
>> aerogear-dev mailing list
>> aerogear-dev(a)lists.jboss.org
>> https://lists.jboss.org/mailman/listinfo/aerogear-dev
>
> --
> abstractj
>
>
> _______________________________________________
> aerogear-dev mailing list
> aerogear-dev(a)lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/aerogear-dev
_______________________________________________
aerogear-dev mailing list
aerogear-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/aerogear-dev