Lightweight Identity


Johannes Ernst contacted me today to tell me about Lightweight Identity (LID). Coincidentally, I'd seen it on Jamie Lewis' blog last week and had it on my list of things to write about (which is essentially equal to my list of things I want to know more about). I first met Johannes in May 2003 at a Jupiter conference on blogging in business. Johannes' company, NetMesh developed LID as a simple, easy-to-use, decentralized way to create identities. LID has a few features which will appeal to many:

  • Identities are URLs (no new namespace)
  • You control the URL and what's there (completely decentralized)
  • Built on standards including vCard, FOAF, XPath, and GPG

Johannes argues why he thinks LID obeys laws of identity. This is good because it will give some structure to Kim's arguments and point out how multiple, different systems might all obey those laws. They represent minimal rule-sets (things you cannot do), not maximal rule-sets (things you must do).

There are several responses from Dave Weinberger, Scott Loftenness, and Eric Sigler. These are all interesting parts of the conversation, but I think miss the point to some degree. The question in my mind is not whether or not LID a good system for storing identities and producing, upon request, identity information. History has shown us that lots of systems can be used as long as they're good enough and LID, along with SXIP, Identity Commons and others are probably good enough on those terms.

The question for me is one of trust, or as Kim likes to call it "recognition." When I use LID to retrieve Johannes' attributes, how do I know that they're OK? Even if I believe that they are exactly as he asserted them (i.e. I believe Johannes is tell me what his address is), how do I trust his assertions? In the real world, I may be having a business meeting with you and you give me a business card. For purposes of getting in touch with you, I believe your assertions because the stakes aren't that high. On the other hand, I may want to know, with some degree of assurance, what your name is. I'd ask for your driver's license. In that case, you're not asserting a value for your name, the government is. Or at least asserting that the person in the picture has a particular name, address, etc. That's the missing piece. LID let's me build business cards, not credentials.

For many things, that's OK. For others, its not. The problem is that making assertions that are trusted by others takes time and carries risk. Risk costs money.

So, going back to the physical world, suppose you apply for a credit card. You fill out a form, asserting a lot of things about yourself. LID could surely do that. Now you send in your form to the credit card company and they verify your assertions, primarily by doing a credit check. There are several companies that collect credit histories and provide credit scores to anyone willing to pay. Those credit scores, of course, are not assertions about a person, but about an identifier (the SSN in the US). Using that score and other information, the credit card company evaluates the risk and issues credit (or not). They pay money to reduce their risk. The credit history company charges, in part, to cover their risk (since they're liable for providing good information).

This presents problems and opportunity. Digital certificates (a way of transferring trust) cost money in part to cover the risk that digital certificate providers incur when they issue a certificate. That makes digital certificates useful in only certain places. It also means that some people will be willing to pay to reduce risk associated with digital identity. There's businesses to be built there.

The interesting thing is that "trust" or "recognition" is about relationships. This points to a way out of siloed identities. It doesn't matter as much that I've got one identity at Amazon and another at BYU as long as there is a mechanism for asserting a relationship between these (i.e. that they both refer to the same person) that can be trusted and an infrastructure that Amazon and BYU can build upon to leverage that relationship.

Randy Gordon wrote to me a few days ago and was talking about the mathematics of identity. In particular, he referenced this Ph.D. thesis from Japan by Tadao Ishii on neoclassical logics with identity connectives. Randy believes, and I agree, there there is room to formalize some of the identity discussion in a language for making identity assertions like the ones I mentioned in the last paragraph. That language could be the basis for building a system of trusted relationships between the referents of an entity.