To: speyer@mcc.com
In-Reply-To: speyer@mcc.com's message of Mon, 16 May 1994 12:05:41 -0700 <94May16.120556pdt.2768@golden.parc.xerox.com>
Subject: Re: URN Requirements
From: Larry Masinter <masinter@parc.xerox.com>
Message-Id: <94May16.130645pdt.2760@golden.parc.xerox.com>
Date: Mon, 16 May 1994 13:06:38 PDT
>>>>> On Mon, 16 May 1994 12:05:41 -0700, speyer@mcc.com said:
> This concept stabs right at the heart of whither there be or not be URNs
> in reality. The concept of "universal" identifier has been tried umpteen
> times (e.g., OIDs in OO parlance) and isn't workable. It requires a level
> of consistency that isn't practical to maintain across the Internet.
I disagree strongly that there is no workable or practical method for
uniformly and universally naming Internet resources. I think the
general problem of making OIDs in OO systems has stronger requirements
for availability and performance than we've claimed for URN
resolution, however.
Of course, the proof of this will be in the implementation, but you
seem to be implying that we shouldn't even try.
> What is implementable is unique identification within a given context.
> A name may be hierarchical which means it may be partially resolved
> in one context and then partially resolved in another context, until
> fully resolved. One reference to an approach that does this is:
> ftp://ftp.einet.net/iceimt/papers/speyer.ps
It's nice that you've published a solution to a different problem, but
it's hard to see what it brings except perhaps to serve as background
material. The referenced paper does nothing to say why 'universal
naming' is hard, other than to assert that it is.
Larry
p.s. If you publish an application/postscript URL, please make sure
you generate the pages "first page first"; for those who prefer to use
ghostview or pageview to preview such things or who have postscript
printers that actually stack the paper in the right order, the
reverse-page order really interferes with easy reading.