RE: [xri] human readability/usability in XRIs

["Drummond Reed" <drummond.reed@onename.com>]

Mike, good point. This is a huge topic (and it's very late for me), so
I'll just provide a perspective to get discussion started.

The requirement is that XRI syntax support identifiers that can be
"optimized for human readability, memorability, and usability". There
are two ways this could be interpreted. The first is that there is a
form of expressing an XRI that, like many HTTP URIs today, is relative
easy for humans to make sense of, similar to what you illustrate below.

The second one is that fully human-optimized XRIs can be dramatically
simpler than most URIs today because they include as little syntax or
metadata as possible without being ambiguous.

The global context symbols in the strawman were originally developed in
XNS syntax to fulfill this second approach. They are an answer to the
challenge: "What's the smallest amount of metadata that can be added to
a natural language word or phrase to make it unambiguously
machine-understandable and resolvable?"

The three global context symbols - "+" for generic reassignable
identifiers, "@" for proprietary reassignable identifiers, and "=" for
personal reassignable identifiers - were the product of several years
worth of boiling down the answer to this question. They let you
distinguish, for example, between the generic concept of "Shakespeare
the writer" (+Shakespeare), the trademark "Shakespeare the company"
(@Shakespeare), and the "person who goes by the personal name of
Shakespeare" (=Shakespeare).

Most importantly, when it comes to human usability, the expectation is
that the same way "www." is recognized by many resolvers and text
editors today (including the editor I'm typing this message into), the
"+", "@", and "=" global symbols will be recognized by XRI-aware
resolvers and text editors without needing an "xri://" prefix. So you
could type:

+florists
@ABC Company
=Drummond Reed

directly into the addess bar of any XRI-aware device or application
(browser, email client, instant messaging client, PDA, text-enabled cell
phone, etc.) and get appropriate resolution (assuming you had the
necessary access permissions).

Finally, the real power comes through the ability to combine these three
types of identifiers to quickly and uniformly establish what otherwise
can be very difficult issues of context for resolution or searching. For
example:

@ABC Company +florists

would search for only the set of organizations that use the identifier
"ABC Company" and who also identify themselves as florists, and

@ABC Company +florists +London

would further narrow such a search to locations in London.

Of course this raises lots of questions about: search vs. resolution
mechanisms, global uniqueness, the global registries needed to enable
this functionality, privacy-enabled resolution, how/if generic
identifiers are registered, and so on. These are in fact the topics
around which XNSORG was formed (see www.xns.org) and in which OneName
(the company I work for) has proprietary interests (see
http://www.xns.org/pages/licenses-and-agreements.html).

It's a deep, rich topic, so let the discussion begin and I and others
who have worked on XNS will provide all the background and examples we
can.

Best,

=Drummond 

-----Original Message-----
From: Lindelsee, Mike [mailto:mlindels@visa.com]
Sent: Tuesday, April 22, 2003 12:50 PM
To: 'xri@lists.oasis-open.org'
Subject: [xri] human readability/usability in XRIs

Hi All,

I've been thinking about what we mean by human readability/usability and
it is clear that we have different opinions as to what it means.  For
instance, to me it means XRIs that are composed of characters that CAN
be read and understood by human beings (not that it is necessarily easy
to do so).  For instance,

   xri://(http://root.cross-reference.com).leaf/local/part

(to steal an example from the strawman) is something that I consider
both human readable and usable and

  urn:xri://.A4B7.14E8.CA62/.(urn:xri://.DA23.7F2B/.97).7.2

is not (less so in any case).

It was mentioned in the thread on versioning that using cross references
as version tags might not be considered human readable.  I personally
don't think that is the case.  I also find the global contexts defined
in the strawman to make XRIs look more complicated (and to my way of
thinking, less human readable).  Are there good examples that lead to
needing to SYNTACTICALLY distinguish between XRIs that identify
individuals, organizations and concepts?

Mike