What is the purpose of nameidentifier claim?

Anthony Serdyukov picture Anthony Serdyukov · Apr 28, 2011 · Viewed 33.7k times · Source

What the claim of type http://schemas.xmlsoap.org/ws/2005/05/identity/claims/nameidentifier should be used for?

This is the main question, and here are additional ones.

How does it differ from http://schemas.xmlsoap.org/ws/2005/05/identity/claims/name claim?

Is it permanent for particular user as opposed to name claim?

Is it globally-scoped or IdP-scoped?

Answer

EBarr picture EBarr · Jun 21, 2012

Name, is just that a name. If we're talking person, think "Eric"; a server "file01".

A NameIdentifier is the ID for an object. Turning back to our person object, Eric's UserID might be 435 in your database. For the server the Identifier could be something like a FQDN or a SID.

According to this post, apparently Name Identifier was a SAML 1.1 property, and is being supplanted by NameID in SAML 2.0.

Unique or Not?

I wanted to address @Jason's comment and @nzpcmad's post. I don't see uniqueness as a clear cut requirement. The question is tagged but the schema referenced is owned by OASIS. So those are the two parties interpretations we need to balance.

Microsoft's stance for ADFS is clearly that there is a unique requirement. We see that in the "The Role of Claims" article. No doubt ADFS casts a big shadow, but this seems like an implementation detail.

Looking at the SAML 1.1 spec, however, I see no such assertion. The closest we get in section 2.4.2.2 of spec is:

The element specifies a subject by a combination of a name qualifier, a name, and a format. The element has the following attributes:
...
NameQualifier[optional] The security or administrative domain that qualifies the name of the subject. This attribute provides a means to federate names from disparate user stores without collision.

The text of the spec tells me that I need to be able to find a person using a combination of the three attributes, but it makes no assertion as to uniqueness. Couldn't I have two entries that point to the same user? Seems so. Moreover, wouldn't' the spec indicate the NameQualifier attribute was required in cases where NameIdentifier was insufficient to uniquely identify the name?

So what's this all lead to?

  • Be careful, unqiue is likely safer.
  • Dig into your providers stance on the topic.