[OpenID] human-memorable namespaces (was RE: Identityconceptsexplained by cartoon ducks)

Drummond Reed drummond.reed at cordance.net
Sun Jan 7 01:21:10 UTC 2007

>>Drummond Reed wrote:
>> The principle is the same -- you've reduced it to a set of semantic
>> But with DNS or URI syntax, look at the number of semantic units you have
>> and the number of ways to put them together:
>> 	john.alford.smith.domain.tld
>> 	domain.tld/john/alford/smith
>> 	domain.tld/john.alford.smith
>> 	john.alford.smith at domain.tld
>> Each of these involves five semantic units, two of which (domain and tld)
>> don't have anything directly to do with the subject.
>> By contrast a global personal i-name with equal expressive power has just
>> three semantic units, all of which are directly associated with the
>> 	=john.alford.smith
>You're comparing a third-level domain with a second-level XRI. That's 
>hardly fair.

Well, it is fair from the standpoint that XRI syntax is designed to
accommodate human semantic units (dot-delimited segments) inside
machine-semantic units (XRI subsegments), while DNS is not. Again, the XRI
TC suffered through a year-long process and version rev to add this feature
(among others), so it's hardly a trivial point.

>john.alford.smith.name and =john.alford.smith are roughly equivalent. 
>The benefit here is merely that you've replaced the ".name" with an 
>equals sign. Both have four "semantic units".

Yes, good point, Martin. As a DNS TLD namespace, .name is the closest
equivalent to the XRI = namespace.

>john.alford.smith.domain.tld would be more comparable to (for example) 
>@domain*john.alford.smith. Again, both of these have *five* of your 
>"semantic units", except that in this case a large number of domain 
>namespaces (.com, .org, .net and national variants on these) have been 
>collapsed into a single namespace identified by "@".
>You can't really claim that the global context symbol is not a semantic 
>unit while counting the top-level domain as one. I put it to you that 
>you are attempting to mislead people.

I honestly didn't mean to mislead anyone. I'm obviously waaaaay to close to
XRI syntax due to my years of work with it (that's why I value the feedback
on this list so much).

But in my defense, in identifier specifications like URI (RFC 3896), IRI
(RFC 3897), and XRI, there's a very clear line between syntax and semantics.
Every syntactic delimiter is unambigously defined in the ABNF. In an XML
sense, they are the "tags". Every other character in the identifer is
semantic, i.e., they are the "content" contained by the tags.

XRI architecture was very deliberate in creating five logical syntactical
characters, refered to as global context symbols (=, @, !, +, and $ -- from
among the very limited choices that URI syntax gave us) as delimiters for
the purpose of signifying the global context of an identifier. You could say
that these are "semantic units" insofar as they impart meaning to the
identifier. But in the architecture they are in fact syntactic characters,
i.e., they can't contain any "content" the way DNS TLDs can. And that means
that they always signify the same thing, vs. the many different types of
semantics that are carried in different DNS TLD namespaces (two-letter .xx
domains for countries; .gov for US governament, .museum for museums, .pro
for professionals, .travel for travel industry, etc.)

Again, I have no beef to pick with DNS. It's an amazing infrastructure that
is nothing short of miraculous. I'm as big a supporter of DNS as I am of
XRI. And of URIs. The only reason I'm here today is that they couldn't do
everything we needed to do for the persistent identity and trusted data
sharing applications that I started working on 15 years ago. So that's why a
whole bunch of people got together and created what's now XRI.

(As a final note, thanks to those on the list for tolerating all this
traffic about XRI stuff from those of us working on that piece of OpenID
infrastructure. See my next note for a suggestion about a separate list.)


More information about the general mailing list