Message-Id: <199403032145.QAA25047@wilma.cs.utk.edu>
From: Keith Moore <moore@cs.utk.edu>
To: Larry Masinter <masinter@parc.xerox.com>
Subject: Re: allowable variations in URNs
In-Reply-To: Your message of "Thu, 03 Mar 1994 12:38:39 PST."
<94Mar3.123841pst.2732@golden.parc.xerox.com>
Date: Thu, 03 Mar 1994 16:45:37 -0500
> >1)Two URN's are `the same URN' only if they are spelled the same.
> > In particular, URNs are case sensitive, have no
> > optional parts that default, and have no alternative encodings.
>
> I had a private message about case sensitivity in URNs. It made me
> think that in the interest of human transcribability that URNs should
> be `case insentitive' (E.g., that if someone types one with upper
> case, it should be translated to a lower case), and possibly even that
> some of the punctuation marks should be optional (e.g., if ISBN
> numbers appear with or without the hyphens.)
It's difficult to say how far we should go in the direction of transcribability
vs. being able to grandfather in other naming systems. Should we restrict to
just letters/digits and have all other characters be considered optional? Should
we group the letters into five-letter groups (separated by -, say) so they can
be transcribed by average humans?
Also, if we end up defining LIFNs separately from URNs, the requirements are
different for the two (since presumably LIFNs aren't meant to be passed around
on the backs of envelopes).
IMPORTANT NOTE:
The WG should NOT spend a lot of time discussing this. If the transcribability
issue is really important than someone who understands human factors should go
off and do some research about it and see what works, and report back.
Otherwise we will just have a bunch of people speculating about something in
which they don't have much expertise.
> This is an area where two requirements (simple comparison method, vs.
> human transcribability) might be in conflict, and the actual
> specification should try to find a reasonable compromise.
The comparison method is still simple. Just always fold URNs to "canoncal form"
(a single case, with optional chars removed) before storing in a database or
comparing them with other URNs.
(But some implementations will botch this anyway...just like some machines
require the name of a machine in .rhosts to be spelled exactly the same way
as the returned by DNS or YP.)
Keith