PersistentDomains 16 KB

Raw Blame History Permalink

<html xmlns="http://www.w3.org/1999/xhtml">
  <head>
    <meta name="generator" content=
    "HTML Tidy for Mac OS X (vers 31 October 2006 - Apple Inc. build 13), see www.w3.org" />
    <title></title>
    <meta content="text/html; charset=us-ascii" http-equiv=
    "Content-Type" />
    <link href="di.css" rel="stylesheet" type="text/css" />
  </head>
  <body lang="en" xml:lang="en">
    <img align="left" alt="W3C" border="0" hspace="0" src=
    "http://www.w3.org/Icons/WWW/w3c-home" />
    <p>
      Things get written under Design Issues when I have expressed
      them several times in various contexts. This idea hit that
      criterion in a W3C staff meeting in October 2000, as I had
      suggested it in various contexts and various times, including
      backstage to Ether Dyson at the Harvard Internet Society
      conference.
    </p>
    <h4>
      Persistent Domains - Strawman ideas on Web Architecture
    </h4>
    <p>
      Up to <a href="Overview">Design Issues</a>.
    </p>
    <hr />
    <h1>
      Persistent Domains
    </h1>
    <p>
      This is a proposal to address the problems with the
      persistence of HTTP URIs. It introduces the concept of a
      datestamped domain name with associated rights and
      obligations of ownership.
    </p>
    <h2>
      Introduction
    </h2>
    <p>
      &nbsp;The problems of the lack of persistence of URIs lead to
      many well-known problems, including
    </p>
    <ul>
      <li>user frustration and social dysfunction&nbsp;with Error
      404 messages; and
      </li>
      <li>(worse) the dereferencing of a URI using HTTP/DNS leading
      to completely different resource to that intended by the
      referring person.
      </li>
    </ul>
    <p>
      &nbsp;A&nbsp;second-order symptom of the problem is that
      there is a constant stream of proposals for new URI schemes
      with different, incompatible, name-lookup technology. These
      are often made with less attention to the real social issues
      surrounding persistence than HTTP/DNS, but propose to be
      systems otherwise similar except in having greater
      persistence.
    </p>
    <h2>
      The analysis
    </h2>
    <p>
      The persistence of HTTP URIs can be factored into two issues:
    </p>
    <ol>
      <li>The persistence of the opaque string which follows
      the&nbsp;domain name, and
      </li>
      <li>the persistence of the domain name&nbsp; itself.
      </li>
    </ol>
    <p>
      The first of these is an&nbsp;issues is of course under the
      control of the domain owner.&nbsp; This, combined with a
      dearth of tools which help one run a web server with
      persistent URIs, has led to a vastly varying level of
      persistence.&nbsp; Some sites understand and construct their
      URIs carefully, while some obviously have not thought about
      the problem and end up changing URIs out of thoughtlessness
      rather than malice. I have summarized some aspects of this
      problem in "<a href=
      "http://www.w3.org/Provider/Style/URI"><em>Cool URIs Don't
      Change</em></a>" .&nbsp; This relies on publishers making an
      institutional commitment to persistence.&nbsp; I have tried
      to lead the way with W3C's draft <a href=
      "http://www.w3.org/Consortium/Persistence">Persistence
      Policy</a>. However, this only addresses the first element.
    </p>
    <p>
      The second&nbsp;issue is summarized by ICANN chair Ester
      Dyson's line, <em>You don't buy domain names, you rent
      them</em>. This single&nbsp;meme instantly undermines any
      public expectation that domain names should be persistent. In
      principle, if&nbsp; <a href=
      "http://www.example.com/downloads">http://www.example.com/downloads</a>
      today points to example.com's download page, tomorrow, it
      could point to anything as defined by a company which has
      bought, acquired though law suit or accidental expiry the
      domain name "example.com".&nbsp;&nbsp;&nbsp;In fact, however,
      huge numbers of links are being made with HTTP. The planet's
      investment in domain names in references is huge. The
      technology&nbsp;which actually is involved
      in&nbsp;dereferencing such identifiers has become more and
      more&nbsp;sophisticated.&nbsp; In practice, also,
      the&nbsp;legal weight&nbsp; behind a&nbsp;significant
      organization's ownership of a domain name is
      considerable.&nbsp; No one would dream that a legal battle
      <em>microsoft.com</em> or <em>ietf.org</em> would be lost
      to&nbsp;some&nbsp;sneaky entrepreneur.&nbsp; Through the rift
      between trademark law and domain names is a problem, there in
      fact is strong legal support for an organization's ability to
      keep its name. But is this enough?
    </p>
    <h2>
      The Solution
    </h2>
    <p>
      I think we can do better. We can do better, if this scheme
      works out, on both issues.&nbsp; To tackle the second, we can
      create domain names which are allocated once and once only,
      which are bought, not rented. We can simultaneously set
      expectations that such data will endure, that names will not
      be reused, and that information will be available even after
      organizations involved have disappeared.&nbsp; The trick is
      the same one as used in W3C's datespace URIs (such as
      <a href="http://www.w3.org/2000/01/sw">http://www.w3.org/2000/01/sw</a>)
      and British car registration plates: to date code them. For
      example, let us create a series of top-level
      <strong>persistent domains</strong> y2000, y2001 and so on.
      One would only be able to acquire a .y2000 domain name&nbsp;
      during the year 2000. Once acquired one would have it
      forever. In fact, the domain name would correspond forever to
      the information published in it.&nbsp;
    </p>
    <p>
      We could take the precaution of inserting a few rules for
      sanity here.&nbsp; There should be some combination of due
      diligence to ensure you are not infringing trademark when
      applying for an alphanumeric name. I would like to put a
      limit on the number of persistent domains any company could
      own - or at least put those who have n in the queue ahead of
      those who have k&lt;n.&nbsp; There could be a convention that
      if you are happy with a random numeric domain 6872364.y2000
      you can have one immediately and automatically.&nbsp; It
      would be great to put some constraint on sitting on a domain
      without using it, and maybe on transfer of domain ownership.
    </p>
    <p>
      To tackle the first issue, an organization wishing to enter
      the scheme makes necessarily a few commitments. One is that
      it must partake in some cooperating mirroring scheme in which
      other organizations or commercial services take on running
      mirrors of the site.&nbsp; I can imagining this working, for
      example, as a for-pay service for consumers, or a mirror-ring
      system for academic institutes. There must be some form of
      contract which, in the event of the original publisher of the
      information coming to a voluntary or involuntary demise, the
      mirror sites will continue the service.&nbsp; The documents
      then enter an <strong>archive state</strong> in which the
      original publisher loses authority to evolve the domain and
      the public gain rights of access. The actual contractual
      arrangements will in fact have to be skillfully set up. For
      example, there will be some information which will be just so
      uninteresting that no one will be prepared to pay for it in
      the long run, but there must be some way to give serious
      archival institutions (the major national libraries for
      example) the right to take over an archive for the public
      good.&nbsp; However, it would be best to start with simple
      conditions, but allow them to be modified with experience
      (real and&nbsp;imagined - <em>gedanken</em>) of the
      system.&nbsp; Other interesting things which come to
      mind&nbsp;include the mirroring of confidential access
      controlled documents with a 30 year timeout on the
      confidentiality.
    </p>
    <p>
      I don't know how best to enforce that a URI is never
      reallocated to something else by a publisher. Actually, I
      don't think it will be a long term problem, as the tools will
      be made such that it is not a function. ("Rename: command not
      known"!).&nbsp; Obviously&nbsp;re-use would cause a big mess,
      as caches across the globe would run out of sync, and
      assumptions made by mirrors and proxies would become
      invalid.&nbsp; Perhaps a suitable disposition for a domain
      whose publisher consistently flaunts the rules would be for
      it to be more or less automatically declared dead, and for it
      to pass into archive state just as though the organization
      had passed away. The organization can then ask for a new
      domain and start again - at least the first time.
    </p>
    <p>
      A clarification of what re-using a URI means.&nbsp; As
      I&nbsp;have pointed out many times, most URIs are <a href=
      "http://www.w3.org/DesignIssues/Generic">Generic URIs</a>.
      They refer to, not a fixed set of bits, but a conceptual
      resource whose representation may vary with (for example)
      time, technology, language, and so on.&nbsp; This is fine,
      even in a persistent domain. It is in my opinion important to
      distinguish (and ideally independently identify with separate
      URIs) generic resources and any specific representation in
      use in a specific case.&nbsp; The contractual obligations of
      ownership of a persistent domain could usefully include the
      provision of such metadata, and the allocation of a
      persistent URI to the actual (mime-type, octet-string)
      specific entity.
    </p>
    <p>
      Obligations of persistence&nbsp;would have to
      apply&nbsp;across the transfer of the persistent domain. The
      persistent domain would be property, with (like land) rights
      and obligations. Transfer of the domain name would carry
      within it both.&nbsp; Fleet Bank would not be able to buy
      BankBoston and simply drop support for documents
      bankboston.y2000 - without&nbsp;all public documents falling
      into archive state.
    </p>
    <p>
      Another clarification.&nbsp; A document in archive state has
      a certain general license to mirror and reproduce in
      unmutilated form, but the intellectual property rights remain
      as ever with the author. A very tempting rule (which I am not
      convinced of yet) is that once the original domain owner has
      defaulted on publishing the material, that it acquires some
      limited redistribution license<a href=
      "#Print"><sup>*</sup></a>.&nbsp; This is the Web equivalent
      of an implicit right to photocopy any book out of print: the
      publisher has deigned not to make copies, and it impedes the
      operation of society to prevent this erstwhile public
      information from being accessible.
    </p>
    <h2>
      Summary&nbsp;
    </h2>
    <p>
      Here is a collection of simplified rules which would form the
      protocol of persistent domains.
    </p>
    <ol>
      <li>Domain names *.y2000 only allocated in 2000, and so on;
      </li>
      <li>Some level of trademark due diligence before registration
      for alpha numeric names
      </li>
      <li>Random numeric names&nbsp;&nbsp; 123678.y2000 issued
      immediately but not transferable
      </li>
      <li>A limit on the number of persistent domain per
      organization would be useful
      </li>
      <li>Domain names owned for life and persist forever
      </li>
      <li>URIs may be live or frozen but not reused arbitrarily
      </li>
      <li>Implicit irrevocable license for 3rd parties to mirror
      public info now and after death
      </li>
      <li>The registry(-ar - whatever is the controlling authority,
      I forget the terminology) would be a neutral non-profit
      cooperative.
      </li>
      <li>ICANN's delegation of .y[1-9]* would be irrevocable in
      all time.
      </li>
      <li>Anyone using it would pay me $1 (just kidding! :-)
      </li>
    </ol>
    <p>
      Provided these are put together with sufficient care, the
      system should run itself in such a way&nbsp;as to preserve
      our information world for posterity,&nbsp;be it represented
      by a consumer in&nbsp;a decade searching for instructions
      from the now-defunct manufacturer of an appliance, or by a
      historian&nbsp;in&nbsp;a millennium&nbsp;trying to figure out
      what on earth made us all tick way back then.
    </p>
    <hr />
    <p>
      &nbsp;
    </p>
    <address>
      <p>
        Tim Berners-Lee
      </p><br />
      <p>
        First Written: 2000/10/04
      </p>
    </address>
    <p>
      &nbsp;
    </p>
    <h2>
      <a name="Footnotes" id="Footnotes">Footnotes</a>
    </h2>
    <h3 id="Print">
      Out of Print Books
    </h3>
    <p>
      Joseph Reagle points out a <a href=
      "http://www.usg.edu/admin/legal/copyright/copy.html#part2a4">passage</a>
      in <em>[University System of Georgia]</em> <em>Regents Guide
      to Understanding Copyright and Educational Fair Use</em>
      which explains:
    </p>
    <blockquote>
      <p>
        4. Out-of-Print-Book SCENARIO D: A library has a book that
        is out of print and unavailable. The book is an important
        one in the professor's field that she needs for her
        research. QUESTION a: May the professor copy the book for
        her files? ANSWER: Yes. This is another example of personal
        use. If one engages in the fair use analysis, one finds
        that: (1) the purpose of the use is educational versus
        commercial; (2) the professor is using the book, a creative
        work, for research purposes; (3) copying the entire book
        would normally exceed the bounds of fair use, however,
        since the book is out of print and no longer available from
        any other source, the copying is acceptable; (4) finally,
        the copying will have no impact on the market for the book
        because the book is no longer available from any other
        source. QUESTION b: Using the same facts as explained in
        SCENARIO D, could the professor copy the book and place the
        book on reserve in the library? Could the professor scan
        the book into her computer and place the book onto the
        World Wide Web? ANSWER: If the professor placed the book on
        reserve in the library, the use would be considered a fair
        use. However, if the professor placed the book on the Web,
        then the use is not a fair use. Placement on the Web allows
        unlimited access to the book. This would affect the
        copyright holder's public distribution of the book. See
        SCENARIO R, SCENARIO T, and SCENARIO U.
      </p>
    </blockquote>
    <p>
      Joseph points out that the courts would not necessarily
      accept a parallel between Web and paper distrbution. My
      purpose in drawing a parallel was to explain the intent, not
      to suggest that either action was in fact admissable under
      any particular jurisdiction.
    </p>
    <h3 id="Note">
      Note for readers after 2100
    </h3>
    <blockquote>
      <p>
        Note for readers after 2100.&nbsp; To understand this
        document you must first understand the situation which
        pertained at the dawn of M3. The anarchic chaos which
        reigned over the embryonic Web is often hard for students
        to grasp, and the fact that today's Web grew from these
        chaotic beginnings is a constant marvel. The reader would
        do well to consult McKinley's "Chaos and
        Fortune:Civilization from the Robber Barons to Domain
        Sharks" [Time-Microsoft-Murdock, 2057] for a simple sketch,
        and Deaton and Plim's "Social Anthropology of the
        pre-fusion Human" [UNArchivePress, 2068] for a&nbsp;more
        detailed analysis.
      </p>
    </blockquote>
    <p>
      &nbsp;
    </p>
  </body>
</html>