Using an ID service
What is an ID service ?​
An ID service, as we chose to call it, is a process that allows for generating a unique browser identifier and enhancing the Snowplow tracking capabilities in environments where the Intelligent Tracking Prevention (ITP) feature is enabled such as on iOS browsers (Safari, Chrome, Firefox) and desktop Safari.
Why you would want to use an ID service ?​
As of Safari 16.4 released in April 2023, Safari sets the lifetime of server-set first-party cookies to a maximum of 7 days in the following cases:
- The server setting the cookie is behind a CNAME that resolves (at any point) to a host that is third-party to the website the user is currently browsing.
- The server setting the cookie is set with A/AAAA records that resolve to an IP address (IP4 or IP6) where the first half of the address does not match the first half of the IP address for the server on the website the user is currently browsing. (e.g. 203.52.1.2 and 203.52.56.22 are okay, 201.55.1.2 is not).
This greatly limits the effectiveness of tracking a customer journey where users are not regularly returning to your website. Without a strong understanding of your customers, downstream use cases including marketing attribution, product analytics and personalized recommendations are difficult to achieve.
The ID service can help enhance the confidence in persistent browser identifiers that Snowplow tracking provides and specifically for this guide, the network_userid
.
Developing and deploying an ID service​
An ID service is code that needs to be deployed on and executed from the same IP space that serves the main web document of your application. This is probably the web application system or the CDN in front of the application.
This code has minimal functionality and based on our experience can either be:
- An API endpoint that is part of the main web application.
- A function running in an edge worker. Almost all modern CDNs allow the addition of ‘middleware’ code which runs on requests at edge nodes in front of the client’s application domain.
- A custom middleware based on the customer’s framework. E.g. ExpressJS middleware, Next.js middleware, Play custom action etc. which can run on every document request.
- A low-footprint application with a single endpoint. A Go web server or something along these lines.
Developing the ID service code​
The responsibilities of this service are:
- Create a unique identifier (UUID v4) for this browser, set it in a cookie and return it in a
Set-Cookie
response header on a domain accessible by the service at all times. - Increase the expiry for the cookie used as the network_userid identifier (by default, a cookie named sp, configured via collector.config.name) which should have the same value as the first cookie.
The new unique identifier cookie for sake of simplicity in this document will have the name spIdService
.
ID service business logic​
The ID service code should include the following logic:
- If the ID service new identifier cookie already exists on the request, then it should re-set the cookies with the same values and updated expiration for both the
spIdService
andsp
cookies. - IfÂ
spIdService
 does not exist, but the collector'sÂsp
 cookie does exist, then the it should setÂspIdService value = sp cookie value
. This will make sure we keep old identifiers in place and not lose any data. - IfÂ
spIdService
 andÂsp
 are both missing, then it generates a new ID in theÂspIdService
 andsp
cookies with the same unique identifier generation algorithm with the Snowplow pipeline, currently UUID v4. - The HTTP response should have a 200 OK status code but any additional payload is not necessary.
Code examples​
Below we showcase a couple of code samples for ID service API endpoints:
- Next.js TypeScript
- PHP
loading...
loading...
Using the ID service on the Snowplow browser tracker​
When the ID service has been deployed on a system with the same resolved IP as the main document, the tracker can then be configured to orchestrate the required ID service API calls.
This process is opt-in by using the idService
option during tracker initialization:
newTracker('sp1', 'c.customer.com', {
idService: "/id-service-endpoint",
/* ...Rest of the tracker options */
});
When the tracker detects this option it will send an HTTP request during initialization on this endpoint to have the service set the required identifiers before sending any event.