"Best practices" for different kinds of agents

I've been trying to sort out the various kinds of processing
agents that deal with IRIs for which we might want to give
different kinds of "best practices" guidelines, in the sense
of being conservative producers and liberal consumers, in
order to improve reliability of transmission.

a) Agents that take an IRI and produce a visual representation
b) Agents that allow a user to 'copy' an IRI and later 'paste'
  it into another (running text) environment, or otherwise
   inject IRIs into running text.
c) Agents that allow a user to input via some entry method
   some characters and transform these into IRIs.
d) Agents that scan a (running text) context and detect
   IRIs along with their boundaries
e) Agents that attempt to decide whether two IRIs would
   connect with the same resource when resolved.

Some amount of difference between the result of (a) vs. the
result of (b) followed by a display of the running text might
be allowed, e.g., IRIs might be visually presented  (a)
with spaces which, in (b) are turned into %20.

In addition, we might also list 'best practices' for

f) Agents that assign IRIs for accessing resources
   to choose IRIs that robustly survive the deployed
   infrastructure (even parts that don't follow best
   practice), with respect to (a) followed by (c)
   or (b) followed by (d).

Is this a useful analysis? The guidelines and best
practices should be extended for

   IRIs that contain strings in RTL scripts (BIDI IRIs)
   IRIs that contain combining characters, accents


-----Original Message-----
From: public-iri-request@w3.org [mailto:public-iri-request@w3.org] On
Behalf Of Ted Hardie
Sent: Tuesday, May 25, 2010 4:20 PM
To: public-iri@w3.org
Subject: Requirement for internal direction changes (was Re: [bidi]
Re: Special ordering for BIDI URLs)

Forgive my failing memory here, but I thought there had been
discussion
at one of our face-to-face meetings about contexts in which there were
different expectations about script direction within a single context.
One
discussion point that I seem to recall touched on a context in which
the
local language was right-to-left but numbers appearing within the text
using Hindu-Arabic numerals (0-9) were left-to-right.

If that memory is correct, or there are other contexts in which a
right-to-left
and left-to-right context can mix, it seems to me that we cannot use
the simplifying assumption that we can use a marker (either embedded
in the IRI or at the presentation layer) to reverse the whole string.

Am I misremembering this or misinterpreting the consequences?

regards,

Ted Hardie

Received on Wednesday, 26 May 2010 15:46:10 UTC