Re: URNs don't provide permanence

David G. Durand (
Thu, 9 Mar 1995 00:36:22 -0500

Message-Id: <ab82c88a040210046f11@DialupEudora>
Date: Thu, 9 Mar 1995 00:36:22 -0500
From: "David G. Durand" <> (David G. Durand)
Subject: Re: URNs don't provide permanence

>Subject: Re: Ephemeral URLs (fwd)
>From: Larry Masinter <>
>Date: Thu, 2 Mar 1995 19:34:41 PST
>The URN by itself does not supply the infrastructure to promote
>availability, any more than the ISBN number, by itself, provides the
>infrastructure of booksellers, libraries and other organizations that
>allow us to ensure books are available even if the publisher is no
>The URN descriptions so far have not, as far as I can see, explained
>the motivation which will cause such a mechanism to arise and remain
>supported for the length of time we normally associate with the
>availability of physical scholarly artifacts.

Sorry for quoting in full, and then answering in a blob, but the original
message seems short enough:

This cannot, in general, be done.

   As I understand it, the ISBN people are not committed to keeping track
of all ISBNs ever issued, but only track the publishers who assign
identifiers to books. Libraries do keep track of ISBNs, but they decided to
do that for their own motivations, because they are useful, guaranteed
unique identfiers for books, and librarians track many books.

    People will track URNs because they are uniform unique names for useful
data. URLs ar3e definitely worse: they depend on the current locations of
documents. Any attempt to fix this and make them long-lived means comitting
to a URL forwarding architecture at least as complicated as the
infrastructure required for a URN system, at which point the interpretation
of URLs as locators will ceome increasingly irrelevant, as doucments change

   If URNs are based on ISO Public Text Identifiers, or ISBN publisher IDs,
or some other issuing authority-based registry then we've at least handled
the uniqueness problem, we can find out who issued the name, and we've also
hooked onto an existing institution that is likely to survive that is
capable of handing out the identifiers.

   What kind of guarantee could be given that the names will be resolvable
in the future? We can't _require_ anyone to keep track of the information.
What is the penalty mechanism if they toss their index or go under? If an
organization is gone, someone has to pay to remember the information. I
think that we can expect that, just as librarians track ISBNs, they will
track the URNs of electronic resources. But lots of publications fall
through the library system, too: finding good records on pulp-novels of the
50s is hard. Go further back and it just gets worse. Much information is
already irretrievably lost about the victorian "penny dreadfuls," even
though historical research now finds valuable information in this sort of
popular culture. The same thing will happen for URNs perceived not to be of
lasting interest. However, URNs at least provide a sound basis for
retaining that information information.

    Unfortunately, the same technology that makes saving large resource
indexes possible is also making the stuff we might want to save even
larger. Is anyone archiving Usenet news as a whole? I don't think so... but
in 50 years cultural critics will be poring over what samples have remained
to look at the history of early digital culture.

   -- David G. Durand
      Boston University Computer Science