[OpenID] GRDDL and OpenID
James A. Donald
jamesd at echeque.com
Sat Feb 3 03:18:06 UTC 2007
Martin Atkins wrote:
> I believe, based on various blog entries and
> presentations that I've seen, that for many people the
> simplicity of those two lines of HTML is what has
> driven so many early adopters to start using and
> promoting OpenID. I'm wary of anything that makes them
> any less simple. I was wary enough about the need to
> now specify rel="openid.server openid2.provider"
> rather than the single rel value as before.
> What are the benefits of using a URI for a rel over a
> suitably-prefixed, well-documented, memorable string?
In the general case the problem lies in "well
Anyone who wants a new string to support new meanings,
new variations of the protocol, has to get everyone
involved around a table, and they all agree on the new
short and memorable name, and what the new name means.
(It is very hard to get consensus over a mailing list.
So usually you discuss on a mailing list, then hold a
In the early days of the internet, the Internet Assigned
Names and Numbers Authority, the Internet Engineering
Task Force, and the rest of that alphabet soup, were
five guys sitting around a coffee table. It took them
fifteen minutes to agree on something, then they would
wander off to their computers, and in an hour, what they
agreed on, would become reality, and frequently what
they had agreed and made real remained set in stone to
the present day.
And the reason it became set in stone is that as the
internet grew, the various bodies grew and grew,
incorporating more and more conflicting interests, until
finally they suffered total bureaucratic gridlock and
URLs, unlike memorable well documented strings, are
unique and self documenting, thus anyone can produce one
of his own, without the need to build consensus, and
thus without the risk of the gridlock and paralysis that
has afflicted internet governance.
However the same benefit could be obtained without the
cost of long winded URLs by having the following
Anyone who wants to introduce a new string
without consensus may do so, but it has to be
*non* *memorable* - a random twenty four bit
number expressed in hexadecimal beginning with
an alphabetic character, and the documentation
for the string D6B5F73 has to contain the string
"Canonical OpenId Documentation of D6B5F73" so
that it can be found by googling.
Memorable strings require consensus, and their
documentation becomes part of the official
James A. Donald
More information about the general