I'm working on an iPhone app that ideally uses OAuth to communicate with Twitter. I know a lot of people are doing the OAuth workflow inside of their apps using a UIWebView, but I don't agree with that and am going with the Pownce approach.
The problem is, Twitter has this whole scheme for working with desktop apps, using a pin number. When I register my app with Twitter, they have a web form asking me if I'm a desktop or web client. If I choose desktop client, when I try to have the user authorize, I can set the oauth_callback parameter but Twitter will ignore it after authorization and show a pin number. If, on twitter's form, I specify that I'm a web client, it requires me to enter a URL to redirect to after authorization. And, since I'm using an iPhone app-specific url scheme, their web form fails on validation as it only seems to accepts URLs conforming to the HTTP protocol.
So, it seems like I'm stuck - I can't say "desktop" because I don't want to bother with a pin, and I can't say "web" or I can't use an iPhone app URL. Any solution to this?
From your question:
I know a lot of people are doing the OAuth workflow inside of their apps using a UIWebView, but I don't agree with that and am going with the Pownce approach.
The Pownce article suggests that quitting your application and opening Mobile Safari to perform the authentication step is problematic, and that they started receiving bad reviews from users for doing it that way. They also experienced a failure rate of around 40%.
Pownce's solution is to use a UIWebView within your application instead, so I have a feeling you may have misinterpreted their recommendations. That being said, they do label this as a "naive" solution and go on to suggest a bunch of theoretical "ideal" solutions.
Another point you might not realise is that desktop applications (using the "out of band" / pin number method) and web applications need to open the Twitter site in either an embedded or external browser.
So you've got two choices on the iPhone:
Open up twitter.com in a UIWebView, specifying no oauth_callback parameter or oauth_callback=oob to start the pin-based out-of-band flow. The user then needs to copy the pin using the iPhone's copy-paste functionality, manually close the UIWebView, and paste the pin into your application. The pin can then be used converted to an access token.
Do it how everyone else is doing it (UIWebView + custom-uri://foo.bar in the callback parameter).
For obvious reasons, the first option is pretty crap and really only useful on platforms where Twitter is unable to redirect to a custom URI.
A simple solution may be to create an HTTP page that always sends a 301 redirect to your custom URL scheme and then provide that HTTP URL to the twitter web API.
Aside from that, Nathan's answer is very complete.
Here's how I do it: tell Twitter you're a web app, and make up any old HTTP:// URL to satisfy Twitter during registration of your oauth client.
Then in your app, pass the URL you want to in the callback parameter. Twitter (in my experience) uses the one you give it.
You could use an intermediate website for the authentication. Your app creates a unique id (hardware based?) and stores it. It then records it has sent the user for authentication and sends the unique id to your website. It then redirects the user to your website. Your website then sends the user to Twitter using oAuth. The user returns to your website and you mark the unique id as authenticated and store the authentication information. The user restarts the app on the iPhone, it reads it has sent the user for authentication and contacts your website with the unique id - and reads in the authentication information.
Long winded and needs another website, but it should work.
Related
I'm a beginner developer. I am writing an iOS application using Flutter. I need to use an API to identify my user. The provider of that API asks me to send them a redirect URL. From what I understand, when an user identify themselves on my app, a request will be send to the API provider, and once the user is correctly identified, the API provider want to know where to send the user (the redirect URL). But I want the user to stay on my app.
How should I set up that ?
Thanks
The usual way to implement this is via the AppAuth pattern, where the app invokes an ASWebAuthenticationSession window, which is a secure browser. From a UX viewpoint it looks integrated and remains a part of the app.
As a next step, maybe have a look at my iOS AppAuth Basic Tutorial to see what this looks like. The simplest option is a private URI scheme based redirect URL that looks like this:
net.openid.appauthdemo:/oauth2redirect
Once the concepts are understood, and your team has reviewed the user experience etc, the next step would be to implement this behaviour in Flutter.
This is a first time I am trying to make a facebook application/game, and I have couple of questions.
Let's say canvas url is http://mysite/first_app/ and facebook app url is https://apps.facebook.com/first_app
I wonder if visited both url-s should a functionality be the same? I mean does facebook give you more tools and freedom if navigated from https://apps.facebook.com/first_app? Or should both url-s look and do the same thing? Is it possible?
I wonder if visited both urls should a functionality be the same
Of course yes. The difference is just that one is deployed at the facebook's app center and the other at your domain.
Yes, you have more tools available in the canvas ("apps.facebook.com.."), as it will send you the user-id without any user consent. Has the user previously used your app, you will also be able to get their accesstoken and fetch more info about them.
It requires some backend code to decode the signed request parameter, which provides you this stuff.
This example is how php handles signed request:
https://developers.facebook.com/docs/facebook-login/using-login-with-games/
If you're using C#, download the Facebook C# (available through NuGet, the package from Outercurve Foundation) and use this example:
How do I parse a signed request in Facebook C# SDK?
I try to handle the user in Facebook context whenever possible. It adds trust and more tools. Although since the Facebook phone/tablet app (App Store/Google Play) doesn't support Facebook apps at all, I often optimize the app for mobile web use also. I read somewhere that over 50% of daily Facebook visists are done using the phone app, so it might be something to think into your app while developing.
UPDATE:
I can see I have misunderstood part of signed request, as commented by Shadowfax and CBroe. In my app, I use the user id from signed request but of course only when the user has already logged on to the app beforehand. My apologies.
I am trying to create a login system for my iPhone application. My login system is not an ordinary one. Here is what I want the iphone to do.
First of all I will post to a website
"https://www.example.com/login"
my username and password.
If login is achieved (i.e. username and password are correct), it will redirect me to
"https://www.example.com/login/success"
else it will redirect me to
"https://www.example.com/login/retry".
I know it sounds strange but thats the way it works. By the way my application is programmed using Objective-C.
So what I want from the login system is to detect the address of the redirected site and if it is the success one, it will allow me to view the rest of the application.
Can I do that login system on iOS? Do you have any idea about that?
The NSHTTPURLResponse class has a method allHeaderFields that you can use to pull out the Location header. Be aware that header names are case-insensitive.
It sounds like you are trying to use the actual website for authentication. While this may be possible, I would advise against it. You should write a separate web service (aka an API) for the app to interact with. If you aren't familiar with the concept of an API do a quick google search.
For a middleware system with internet (which works inside a set-top box) I want to develop a primitive Facebook interface where users can type their user-names and password, showing their latest notification, messages and other casual stuff on the TV screen by using the recent Facebook Graph API.
This middleware program uses Java ME to run programs (such as this simple facebook app) and it can connect to internet however it doesn't have a real web browser. Without browser it can connect to any url to retrieve the JSON response however I am not sure how to achieve authentication without a real browser.
Under this circumstances, is it possible Facebook authentication? If you think so, what approach would you suggest ?
Thanks
Facebook provides trusted partners with a private Authorization API to get an OAuth 2 token from a username / password.
A more complicated approach would be doing something similar to how Netflix enrolls a device:
device calls server to obtain a Code
device shows code on screen and directs user to go to URL on server and enter Code
server redirects user to Facebook and obtains OAuth token, user told to go back to device
device calls server with Code and obtains OAuth token
device can now make calls directly on behalf of user
According to this documentation on "Desktop Application Authentication" I don't believe your desired result is possible:
Facebook's OAuth implementation does not include explicit desktop application support. However, if your desktop application can embed a Web browser, you can add Facebook support to your application easily using the same OAuth User-Agent Flow used by JavaScript clients.
However, it is clearly possible for certain vendors to do this, since Microsoft's Xbox 360 Facebook application does exactly what you are proposing. I'd be interested to see if anyone has dug up any API for doing this that Facebook doesn't want in their most obvious documentation.
This isn't an answer but I'm trying to do the same thing. Check out this guy's blog which uses another server to proxy the requests:
cory wiles blog
If you figure it out please post a detailed answer here so I can do it to.. :)
I think it is possible though it is pretty complicated and subject to sudden changes of Facebook interface. It might break the agreement between you and Facebook.
What you do is to emulate the Facebook.
One path you have to set up a Facebook application. Once you got the authorisation from user, you can to something with Graph API.
You need to the Facebook log-in process and authorisation process. There are some capturing tools on http/https request and response. Analyse them, both header and body.
Once you know the authorisation mechanism, you can replace it with you own. Everything afterward is on Graph API.
Another path is to emulate Facebook login and message and notification process. Capturing and analysis is needed.
In the past I have used a tool called screen-scraper (full disclosure: I used to work there) to automate logging in to facebook. Basically, it imitates a browser session; it allows you to set session variables (i.e. username, password) which would then be submitted to facebook, just as if the user had submitted them in a browser.
You may not be able to use screen-scraper in your set-top box environment (although it is java-based, so it's possible it would work). Even if it doesn't, you could implement a similar strategy in java, making the HTTP calls a browser would make to load the login page and submit the user's credentials. To keep the user's info safe make sure whatever HTTP client library you use supports HTTPS.
Proxy tools and extensions like Charles, Fiddler2, Firebug, Chrome's dev tools, etc. are helpful in seeing exactly what the browser is sending to the server in requests.
I'm trying to glue information from a web page to an iPhone app that said web page suggests to download. I control both the web page and the downloadable app.
Scenario is like this:
User visits my web page, on which I recognize the user (he may have logged in, and I store his info in a cookie). I then present a link to him to an app in the App Store that he should download for "enhanced experience" of this web service of mine.
Now, when the user launches the downloaded app on his iPhone, I like to re-identify the user who previously visited said web page.
All would be easy if an iPhone app could read Safari's cookies. But it can't.
A somewhat lame solution could be that the web server stores the visitor's IP address and uses that to recognize him once he launches the iPhone app. But that's not reliable.
Another one would be to give the user a token (code) that he needs to remember and then re-enter in the app. Still quite awkward, I think.
Any better suggestions?
Simply put, you can't do this.
One thing you could consider is a custom URL scheme to launch the app. You could send the user an email that uses this custom link. However there's a couple of problems with this:
the user may not have the account that they used to register for your site set up on their iPhone. This might seem unlikely, but say the user signed up for your site 5 years ago with their Hotmail account and they have since switched to Gmail.
it's unlikely that the email would fit into their workflow. They would probably download the app and just launch it by touching the icon instead of clicking a link in a received email.
You could also put the custom URL as a link on your web page, but again, this won't fit into the workflow because they have to go to the App Store app to do the download.
Consider this - if you've got some sort of website that has an authentication step, it's probably a fair bet to say that the user is the type of person who already has an application such as Facebook installed on their iPhone. They are already used to the paradigm of having to enter their credentials into an application despite the fact that they may have already done it in Safari.
If you could read the unique iPhone device ID from javascript on your web page, you could look for that again when the application connected...
But I cannot find any means of reading this from Javascript in Mobile Safari, I thought I'd post in case there is a way now to give you another option to consider.
OK, we found a somewhat working solution: The html code can create a cookie. Later, when the app runs, it can't directly read that cookie, of course (due to the sandboxing of iPhone apps). However, it can connect to the server, then open a http URL pointing to the server and including a unique token that it has gotten from the server beforehand. This leads to launching Safari, accessing the server. The server can now read the aforementioned cookie and finally establish the connection with the help of the token.
Just stumbled over this question and I'm curious if you thought about using a UIWebView.
Where the question is - does UIWebView share cookies with safari?
If it does the rest should be easy.
UIWebView's DON'T share cookies with Safari. So unfortunately that is not an option.