/account-sdk-browser

Schibsted Account SDK for browsers

Primary LanguageJavaScriptMIT LicenseMIT

logo

Build Status Code coverage Snyk

Schibsted account SDK for browsers

Welcome! This is the home of the Schibsted account JavaScript SDK for use by any website that wishes to use Schibsted account to sign up and log in users. Use it to generate URLs for your site's "Log in" button, query the logged-in status of your users, and to check whether they have access to products and subscriptions, etc.

Getting started

  1. Do npm install --save @schibsted/account-sdk-browser
  2. Use this library as you would any other npm module: import { Identity, Monetization, Payment } from '@schibsted/account-sdk-browser'
    1. If you use the CommonJS require syntax and you both want to reduce the size of your JavaScript bundle and don't need all of Identity, Monetization and Payment modules from this SDK — it's possible to require('@schibsted/account-sdk-browser/identity') (note the /identity at the end) or /monetization' or /payment'
  3. Build your site as you prefer. This library uses modern JavaScript syntax (including async/await and other ES2017 and WHATWG features) by default. We recommend that you do any transpilation yourself for the browser versions you need to cater to. See this paragraph for info about our babelified version and info about polyfills.

Upgrading from 2.x

If you already use the 2.x branch of the Schibsted account JS SDK, certain changes will be required to use this version of the SDK. We have chosen what we believe to be a middle ground between "remembering the work done in the old SDK" and "starting fresh". Therefore it is recommended that you read this document in full. But ok, let's present some highlighted differences:

Differences from 2.x

  • Instead of using SPiD.init() for initialization, the new SDK exports three classes; Identity, Monetization and Payment
  • Many features (like logging in) requires a redirectUri parameter — both in the 2.x and 3.x versions of the SDK. An important difference in the new version of our backends, is that we strive to be more compliant with OpenID Connect standards. This means that redirect uris need to match exactly (that is — including the query string). This will be a breaking change for some people, because in the 2.x world, a redirect uri might look like https://site.com in self service, and a login attempt with redirectUri=https://site.com?article=1234 would then be ok because it would only match on domain+path — but not query string. However — this will not work in the 3.x world. OpenID Connect does have a suggestion for how to handle these situations though, which is a parameter called state that you send in addition to the redirectUri. See this paragraph for more information
  • The 'SPiD.' string is removed from the name of all SDK events. So the event that used to be 'SPiD.login' is now just 'login'
  • You don't log in by setting window.location. Instead, you use the login() method on an instance of Identity
  • The JavaScript code in this browser SDK does NOT set any document.cookie = ... by default. There is a function enableVarnishCookie that you can call on an Identity instance. This will enable setting the SP_ID cookie whenever hasSession() is called (though most browsers require that you are on a "real domain" for this to work — so, not localhost). Any other cookie that you need set, you will have to set yourself
  • All functions that used to take callback functions in the 2.x version of the SDK don't do that anymore. The new SDK instead uses promises where it makes sense (often written as async functions). For example Identity.getUser() returns a promise. So, for instance if you used to do this in v2.x:
    identity.hasSession((err, data) => {
        if (err) {
            console.log('Nooo!', err)
        } else {
            console.log('Yay', data)
        }
    });
    Now you should instead do:
    // Either
    identity.hasSession()
        .then(data => console.log('Yay', data))
        .catch(err => console.log('Nooo!', err));
    // ... or if you're using async functions
    try {
        const data = await identity.hasSession();
        console.log('Yay', data);
    } catch (err) {
        console.log('Nooo!', err);
    }
  • Listening to events is still supported, although since many functions return Promises, we expect many users will find the use of Promise results preferable. But for those that prefer the events, it works using a function .on that's compatible with Node's EventEmitter. For example SPiD.event.subscribe('SPiD.login', handler) from 2.x becomes Identity.on('login', handler). Also, the functions .off and .once are supported
  • SPiD URI is gone. There are a handful of ***Url() functions in each of the Identity, Monetization and Payment classes for the relevant flows
  • The new SDK has inline jsdoc documentation that's available here instead of tech docs. These documents will always be up to date with the latest release so make sure to run npm outdated in your project to be notified about any new releases
  • The new UI flows are different than the old ones in that they use the Schibsted account API endpoints just like any other client. For most clients this means absolutely nothing at all, but for some, it's quite important; If you have ever asked our support staff to disable certain API endpoint accesses, there is a chance that you'll encounter problems. For instance, if you've set NO ACCESS on the POST /signup endpoint, users will not be able to sign up to your site using the new flows
  • If you use our session-service, there are certain changes in the response from the session endpoint.
    • The userStatus field makes no sense in the session-service world, since it operates "per-site" (there is one id.site.example domain for each site, so being logged in and connected for that site means the same thing. Also, there is the Identity.isConnected function that's still kept in case people prefer to keep the same logic with and without the session-service).
    • The id field (the one returning a MongoDb identifier like abcdef0123456789abcdd00d) has finally been removed. It's been deprecated for a long time. The numeric userId (legacy) and uuid fields are still present.

Polyfills required for older browsers

This SDK uses modern JavaScript features. If you support older browsers, you should use a tool like babel to transform the JavaScript as needed. However — since certain teams have deployment pipelines where it's difficult to do their own transpilation, we do provide some opt-in es5 files as well:

  1. @schibsted/account-sdk-browser/es5: Include both Identity, Monetization and Payment.
  2. @schibsted/account-sdk-browser/es5/global: Include both Identity, Monetization and Payment. In addition, add them as variables to the global window object.
  3. @schibsted/account-sdk-browser/es5/identity, @schibsted/account-sdk-browser/es5/monetization or @schibsted/account-sdk-browser/es5/payment can be used to only include each class by itself.

But then regardless of whether you use the es5 versions or not, you might need to polyfill certain things that might be missing in the browsers you wish to support. A quick test using IE11 showed that we needed polyfills for Promise, URL, Object.entries, fetch, Number.isFinite and Number.isInteger. If you want any sort of debugging to work (say, if you're passing a function using console.log as a parameter to any SDK function that supports logging), you might also need to polyfill console and console.log (yeah, it's baffling, but a known issue in IE). We added them from polyfill.io like this:

<script src="https://cdn.polyfill.io/v2/polyfill.js?features=Promise,URL,Object.entries,fetch,Number.isFinite,Number.isInteger,console,console.log"></script>

Notes on Apple Intelligent Tracking Prevention (ITP)

or.. how I learned to stop worrying and ❤️ the Schibsted account session service

Right, as of Safari 12, we can't rely on making requests from a site domain to the Schibsted account domain. Safari 12 will possibly partition cookies in such requests in an attempt to protect the user's privacy. While this is good for end-users, it presented a real problem for Schibsted account, since our technique for deciding whether a user is logged in, is to send a request in precisely this manner.

There are two ways to deal with this in Safari:

  1. You can continue with 3rd party requests, but this requires an iframe on that 3rd party domain and it also requires user input in said iframe
  2. You can re-design the system to not use 3rd party requests anymore

We've tried to support both. For the 1st strategy, simply continue using the SDK like before. Be sure to call Identity.login when authenticating, and Identity.hasSession when coming back to your site from Schibsted account. This should pop up our so-called "ITP Dialog". This is the iframe mentioned in point 1 above, and clicking the Continue button in that frame will ensure the hasSession call running inside the iframe is successful. The benefit of this strategy is it requires very little work from you. The drawback is that every time you come back from authentication, the user will have to see this "ITP dialog".

So to work with strategy 2, we have re-designed our platform and introduced what we call the session-service. If your site lives on site.example, you should assign a sub-domain for use with the session-service (we propose id.site.example for production and id-pre.site.example for staging — talk to our Customer Success team regarding how to get this set up). The goal is to have https://id.site.example be a DNS name resolving to our session-service, since this enables us to place a cookie on that domain that indicates that the user is logged in.

When the domain is set up, your client needs to be modified by our Customer Success team to enable using this feature (this is done by them setting session_service_domain = https://id.site.example in SysAdmin).

Finally, and this is where you, the user of account-sdk-browser need a change; When creating an instance of Identity or Monetization, include the property sessionDomain: 'https://id.site.example in the constructor'.

So to sum up:

  1. Prepare a subdomain id.site.example (optionally id-pre.site.example for staging)
  2. Enable the session-service for the client you use on your site
  3. Add the sessionDomain property to the Identity or Monetization constructors

1 and 2 requires communication with us, and 3 is done by you at a time of your choosing. The benefit of this strategy is that we should never need to show any dialog or popup to the user, so it reduces friction. The drawback is the work mentioned above.

Simplified login widget

Implementing this functionality requires that your brand uses session service. With this as a starting point, implementing simplified login in production is relatively straightforward:

  1. Ensure that your site has no site specific terms and conditions in the Schibsted account login flow.
  2. Define rules for when and how often the simplified login prompt should be shown to unique users on your site. How you do this is up to you, but we recommend starting with showing the prompt once per user before potentially increasing this frequency over time.
  3. Set up a function to check if users landing on your domain is logged in to your site.
  4. If the user is not logged-in to your site, call the showSimplifiedLoginWidget function. The showSimplifiedLoginWidget accepts the same params as login function (state is required, it might be string or async function). If the simplified login prompt is to be loaded, showSimplifiedLoginWidget will return true.
  5. Set up a way to store information about which users have been shown the simplified login prompt. How you do this is up to you, but one way is to use localStorage. Use this information to execute on the rules defined in #2.

Example project

There is an example that demonstrates how the SDK can be used. The code is here, and you can see it live here. You have a use-case that we haven't thought of? Ask us to add it by creating an issue.

You can use that code as inspiration or just fork and play with it. The account-sdk-browser NPM module is used for authenticating the user with Schibsted account. Take a look at how the SDK is initialized.

When a user wants to log in to your site, you direct them to a UI flow that is hosted by Schibsted Account. We authenticate the user and redirect them back to your site. This final redirect back to your site is done in accordance with the OAuth2 spec. That means that we pass a code in the query string in that redirect uri. You can use that code on your site backend along with your client credentials (client id & secret) to get an Access Token (AT) and Refresh Token (RT). You don't send the AT (and never ever the RT!) to the browser but rather keep it on the server side and associate it with that particular user session in order to be able to call Schibsted account APIs on behalf of that user.

Events

The SDK fires events when something we deem interesting is happening. For example the Identity class emits some events when the user is logged in or logged out. This SDK uses a familar interface that's very similar to Node's EventEmitter. The most important methods are .on(eventName, listener) (to subscribe to an event) and .off(eventName, listener) (to unsubscribe to an event).

Identity

Let's start with a bit of example code:

Example

import { Identity } from '@schibsted/account-sdk-browser'

const identity = new Identity({
    clientId: '56e9a5d1eee0000000000000',
    redirectUri: 'https://awesomenews.site', // ensure it's listed in selfservice
    env: 'PRE', // Schibsted account env. A url or a special key: 'PRE', 'PRO' or 'PRO_NO'
})

async function whenSiteLoaded() {
    const loginContainer = document.getElementById('login-container')
    if (await identity.isLoggedIn()) {
        const user = await identity.getUser()
        const span = document.createElement('span')
        span.textContent = `Hello ${user.givenName}`
        loginContainer.appendChild(span)
    } else {
        loginContainer.innerHTML = '<button class="login-button">Log in</button>'
    }
}

function userClicksLogIn() {
    identity.login({ state: 'some-random-string-1234-foobar-wonky-pig' })
}

Regarding state

This parameter is an OpenID Connect parameter (described in this paragraph in the spec). It's formatted as an opaque string. This means you can send anything that can be serialized to a string. In practice, we have good experience sending something like a JSON value like a base64-url-encoded value — it's just an easy way to avoid browsers or backends messing with special characters.

But as a trivial example, if you call Identity.login(..) with params redirectUri=https://site.com&state=article%3D1234 — then at the end of the authentication flow, the user will be sent back to your redirectUri, and the state parameter will be forwarded along with the auth code parameter.

It is recommended that you provide a unique identifier as part of the state, to prevent CSRF attacks. For example this can be accomplished by:

  1. Your backend generates random token: 1234abcd, saves it in some tokenCache, and forwards to your browser frontend
  2. Your frontend calls Identity.login with state = base64Urlencode({ token: '1234abcd', article: '1234', ... })
  3. When auth flow completes, the user is redirected back to your site. Then, your backend sees the query parameters code (which it can exchange for OAuth tokens for the user) and state
  4. Your backend can do decodedState = base64Urldecode(query.state) and then verify that its tokenCache.contains(decodedState.token). If that fails, then possibly a CSRF attack was attempted. If successful, remove the token from the tokenCache so the same token can't be used again, and continue to show decodedState.article

Authentication methods

Although Schibsted account abstracts away the details of how the users sign up or log in, it's worth mentioning that your end users have a few ways to log in:

  • Username & password: pretty self-explanatory; users register using an email address and a self-chosen password
  • Passwordless - email: here, the users enter their email address and receive a one-time code that they can use to log in
  • Passwordless - SMS (BETA): similar to the previous method but instead of an email address, they receive the code on their phone as an SMS
  • Multifactor authentication: first client indicates which methods should be preferred, later these will be included (if fulfilled) in AMR claim of IDToken

IMPORTANT: Passwordless using SMS is still in BETA. It's only recommended to use it for testing and experimental purposes for now. Please let us know before using this in production.

The default is username & password. If you wish to use one of the passwordless login methods, the login() function takes an optional parameter called acrValues (Authentication Context Class Reference). The acrValues parameter with multifactor authentication can take following values:

  • otp-email - passwordless authentication using code sent to registered email
  • otp-sms - passwordless authentication using code sent to registered phone number
  • otp - authentication using registered one time code generator (https://tools.ietf.org/html/rfc6238)
  • sms - authentication using SMS code sent to phone number
  • mfa - multifactor authentication, any of otp, sms can be completed by user
  • otp sms - multifactor authentication with both otp and sms methods

The classic way to authenticate a user, is to send them from your site to the Schibsted account domain, let the user authenticate there, and then have us redirect them back to your site. If you prefer, we also provide a popup that you can use. In this method, the authentication happens on a separate popup window and at the end of the auth flow. We recommend that you make the popup send a signal to your main page — using postMessage or something similar — to indicate that the user is logged in. If the popup window fails to open, it'll automatically fall back to the redirect flow. The SDK Example project mentioned above demonstrates how it can work. Again, you can see sdk-example if you want a working example.

Is the user logged in?

Schibsted account relies on browser cookies to determine whether a user is recognized as logged in. The SDK provides functions that can be used to check if the user that's visiting your site is already a Schibsted user or not.

  • Identity#isLoggedIn tells you if the user that is visiting your site is already logged in to Schibsted account or not.
  • Identity#isConnected tells you if the user is connected to your client. A user might have isLoggedIn=true and at the same time isConnected=false if they have logged in to Schibsted account, but not accepted terms and privacy policy for your site.

If you've lately changed your terms & conditions, maybe the user still hasn't accepted them. In that case they are considered not connected. In that case, if they click "Log in" from your site, we will just ask them to accept those terms and redirect them right back to your site.

Logging out

If you want to log the user out of Schibsted account, you can call Identity#logout. This will remove the Schibsted account browser session, and so log the user out of all Schibsted sites in that browser.

On your site backend, it may or may not make sense to remove the access/refresh tokens that you got from Schibsted account. This can simply be achieved by removing it from your session or just deleting the session. At this time, there are no ways to invalidate the tokens so they will not be usable. In the future you might be able to invalidate tokens. This comes in handy if you know that a token is compromised and you don't want them to be usable in the future.

Monetization

The preferred method for checking whether a user has access to a product/subscription is Monetization#hasAccess. It requires using session-service, and supports both Schibsted account productId's and Zuora feature id's.

Example

import { Monetization } from '@schibsted/account-sdk-browser'

const monetization = new Monetization({
    clientId: '56e9a5d1eee0000000000000',
    redirectUri: 'https://awesomenews.site', // ensure it's listed in selfservice
    sessionDomain: 'https://id.aweseome.site', // client-configured session-service domain
    env: 'PRE', // Schibsted account env. A url or a special key: 'PRE', 'PRO' or 'PRO_NO'
});

try {
    // Check if the user has access to a a particular product
    const userId = await identity.getUserId();
    const data = await monetization.hasAccess([productId], userId);
    alert(`User has access to ${productId}? ${data.entitled}`)
} catch (err) {
    alert(`Could not query if the user has access to ${productId} because ${err}`)
}

Legacy methods

These two functions require a parameter sp_id that is obtained from Identity#getSpId asynchronously.

Example

import { Monetization } from '@schibsted/account-sdk-browser'

const monetization = new Monetization({
    clientId: '56e9a5d1eee0000000000000',
    redirectUri: 'https://awesomenews.site', // ensure it's listed in selfservice
    env: 'PRE', // Schibsted account env. A url or a special key: 'PRE', 'PRO' or 'PRO_NO'
})

try {
    // Check if the user has access to a a particular product
    // You need the sp_id parameter that is obtained from an Identity instance
    const sp_id = await identity.getSpId()
    const data = await monetization.hasProduct(productId, sp_id)
    alert(`User has access to ${productId}? ${data.result}`)
} catch (err) {
    alert(`Could not query if the user has access to ${productId} because ${err}`)
}

Payment

This class provides methods for paying with a so-called paylink, buying a product, getting links to pages for redeeming voucher codes, reviewing payment history, and more.

Example

import { Payment } from '@schibsted/account-sdk-browser'

const paymentSDK = new Payment({
    clientId: '56e9a5d1eee0000000000000',
    redirectUri: 'https://awesomenews.site', // ensure it's listed in selfservice
    env: 'PRE', // Schibsted account env. A url or a special key: 'PRE', 'PRO' or 'PRO_NO'
})

// Get the url to paymentSDK with paylink
const paylink = '...'
const paylinkUrl = paymentSDK.purchasePaylinkUrl(paylink)

// Or another example --- pay with paylink in a popup
paymentSDK.payWithPaylink(paylink)

Appendix

Cookies

There are some cookies used by Schibsted account. They should all be considered opaque on the browser side. Nevertheless, here is a short description of them.

  1. The autologin cookie (often called 'the remember-me-cookie'): The cookie name in the production environments is vgs_email, because reasons (on PRE, it is called spid-pre-data). It's a JSON string that's encoded using the standard encodeURIComponent() function and is an object that contains two pieces of information that's important:
    • remember: if set to true, the user chose to be remembered and this means we usually support auto-login (that is, if you call the Schibsted account hassession service, and no session can be found in the session database, it will automatically create a new one for the user so that they don't have to authenticate again. If it is false, it should be interpreted as the user does not want to be automatically logged in to any site when their session expires
    • v: the version number
  2. The session cookies: Cookie names in production environments are identity, and SPID_SE or SPID_NO. It contains:
    • user: an object (if it's missing, a call to hassession will return a 401 with a UserException that says No session found)
      • userId identifies the user. We use this property to compare "old" user with "new" user and fire events that indicate that the user has changed
      • is_logged_in indicates if the user is logged in
    • user_tags: a map that contains some flags about the user; namely:
      • is_logged_in indicates if the user is logged in (this seems to be a duplicate of a property with a similar name in the parent user object)
      • terms: a map of term ids that indicate if they've been accepted by the user.
    • referer (yep, missing the double "rr"..): If this is missing, a call to hassession will return a 401 with a UserException that says No session found.

LICENSE

Copyright (c) 2018 Schibsted Products & Technology AS

Licensed under the MIT License

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.