Well-known destinations

Hi Matthew, I just watched your AC talk about well-known destinations with great interest. It's very close to what we're doing at GS1.

The underlying aim is the same: almost every website has exactly the same pages and functions, whatever the language or layout. The same is true for products you might buy. There's a product information page, instructions, allergen info, recall status, sustainability info and so on. So can we make those links machine-discoverable without worrying about whether it's called "login" or "logon"? 

We're working with the concept of Linksets, as defined in RFC 9264 (https://www.rfc-editor.org/rfc/rfc9264.html), which is a simple data structure that uses link relation types to identify what kind of resource is available at the destination. The link relation types at IANA (https://www.iana.org/assignments/link-relations/link-relations.xhtml) include some useful ones but often lacks what you really want ('home' and 'login' are both absent, for example). OK, there's a process for adding more but you get the feeling that, somehow, it's not for you. That's why we created our own set of link relation types (we just call them link types for short). See https://www.gs1.org/voc/?show=linktypes. I don't claim originality here - it's basically HATEAOS and Linked Data.

As with your work, the idea is that if, for example, you want the certificate of organic status associated with a product (or brand) you can follow links of type 'gs1:certificationInfo' which may be labelled with certificación orgánica, βιολογική πιστοποίηση, 有机认证 or органик гэрчилгээ or any of 100 other labels for humans (thank you Google translate!). The machines can help you find what you want.

We also have a definition of a resolver type that can make use of these. That is, given the URI of a product, you can append "?linkType=gs1:certificationInfo" and be redirected to that destination. Our spec for this is at https://ref.gs1.org/standards/resolver/. We make use of ./well-known to declare that the domain name in the product's URI is a conformant resolver. And the basics of this are included in an ISO/IEC standard I'm shepherding through their process at the moment.

I mention 'product URI' - we're in the process of moving from the familiar 1D barcodes with which we've been associated for more than 50 years, to 2D codes, typically to QR Codes, that contain a URI that themselves contain the same product identifier and more. The number of products with these types of 2D code on them now around the world is well into the hundreds of thousands. The video at https://www.zappar.com/zapvision/ shows how one assistive technology company is making use of our 2D codes in their work.

Link relation types have been around since HTML 3 (with the @rel and @rev attributes). See https://www.w3.org/MarkUp/html3/anchors.html. Mark Nottingham's work on Web Linking (RFC 8288 https://datatracker.ietf.org/doc/html/rfc8288) describes the basic structure that RFC 9264 builds upon.  We've tried hard to re-use those long-established mechanisms to achieve what looks a lot like what you're doing, albeit with a slightly different motivation. 

I wouldn't be so rude as to write the (in)famous line "we should align our standards" which always translates roughly as "you should do what we're doing". However, I can't help think that you have perhaps come up with a very good solution to a real problem that is essentially a round thing with an axle through the middle that turns easily, but without making use of existing wheels. Even if you find the GS1 stuff irrelevant, at least please use the link relation types mechanism/extension mechanism. And consider using the two RFCs I've cited as the basis for your efforts. 

Cheers

Phil


---

Phil Archer
Web Solutions Director, GS1
https://www.gs1.org


https://philarcher.org

+44 (0)7887 767755
@philarcher1