Uniquely identifying products

[Osmose]

Currently, products are identified by their full URL. Several shopping sites offer the same product under different URLs, often for SEO purposes.

There's a few options:

• Per-site custom handling of unique IDs in URLs
• Normalize URLs to remove non-canonical bits
• Extract canonical IDs or URLs from the webpage
• Identify products by a compound key of vendor and product title
• Do nothing, URLs might actually be good enough

[biancadanforth]

We should do something, since currently we are using the URL of the page as the product ID, and that's Category 3+ data (for the MVP, we want to keep everything Category 1/2 per @chuckharmston ), since we will be sending the product ID along with many telemetry events.

We could seed a UUID from some of the product information (I'm not sure if generating one from the URL would be considered Category 3+; I've asked rrayborn).

Chuck gave me some examples of UUID modules:

Python's uuid module lets you do that with uuid.uuid5(uuid.NAMESPACE_URL, 'http://www.whatever.com')
Node does too, I just can't speak for it as personally.
https://www.npmjs.com/package/uuid

[biancadanforth]

Sharing this from my discussion with Osmose earlier over Vidyo:

Using URLs to identify the product locally is not a problem. It’s when that data gets sent away from the client's computer that it becomes a problem.

From a data collection perspective, there’s no way to send a product id that identifies a product across users without something that’s category 3+.

What we can do is use a product id locally and send an alternate identifier (e.g. a primary key that is a UUID or ever-increasing number) to Mozilla. This alternate identifier would allow us to identify a product uniquely from the collection of products for a particular user, but it would prevent us from knowing the exact product or page that the user was on.