Loading ...
Sorry, an error occurred while loading the content.
 

Standardization vs. Interoperability

Expand Messages
  • Tim O'Reilly
    ... Just for the record, when I suggested that the IETF would be a better model for a P2P Working Group than the top down industry consortium model that Intel
    Message 1 of 17 , Oct 16, 2000
      Wes Felter wrote:

      > I'm not really questioning the value of P2P, but I do question whether we
      > know enough about it to start standardizing anything. I don't want to see
      > a huge amount of effort put into taming something that is still growing.

      Just for the record, when I suggested that the IETF would be a better
      model for a P2P Working Group than the top down industry consortium
      model that Intel was pushing, I wasn't suggesting that we're ready for
      standardization yet, or that we get the IETF itself involved (although
      that might be a good idea).

      I do think:

      a) Intel's approach is brain dead. (See
      http://www.oreillynet.com/pub/a/p2p/2000/10/13/working_grp.html.) A
      decentralized anyone can play model *like* the IETF is what we need.

      b) What we really need is just a forum for people to get together to
      start conversations about interoperability. Note that interoperability
      discussions are not the same as standards discussions. To me, the role
      of a p2p working group should be to focus on interoperability.

      c) It seems to me that, for example, if all the distributed computation
      players don't agree on some interoperability standards, it would be
      really easy for MS to wait for the dust to settle, and then just add a
      "generalized spare cycles screen saver" into the OS--after all,
      setiathome has only 1 million users, and they're the biggest of the
      bunch. MS could deploy a gazillion clients just like that.

      The same is true in other p2p areas. Having an interoperability layer
      (analogous to the kinds of de-facto standards that kicked off the web
      revolution) would make things much more interesting.



      --
      Tim O'Reilly @ O'Reilly & Associates, Inc.
      101 Morris Street, Sebastopol, CA 95472
      +1 707-829-0515, FAX +1 707-829-0104
      tim@..., http://www.oreilly.com
    • Adam Rifkin
      ... What do you think of the idea of using The Web as that interoperability layer? Run The Web in both directions -- what Tim Berners-Lee and Dave Winer have
      Message 2 of 17 , Oct 16, 2000
        > Having an interoperability layer (analogous to the kinds of
        > de-facto standards that kicked off the web revolution)
        > would make things much more interesting.

        What do you think of the idea of using The Web as that
        interoperability layer? Run The Web in both directions -- what Tim
        Berners-Lee and Dave Winer have called the "Two-Way Web" -- and we
        get standardization *and* interoperability in one swell foop.

        The world now seems ready for peer-to-peer HTTP networking. Jon
        Udell was ahead of his time...

        http://www.byte.com/column/BYT20000830S0001

        ----
        Adam@...

        There are no successful fully distributed applications out there
        except for Gnutella - in that, it's really a revolution.
        -- Gene Kan, http://www.internetworld.com/article_bot.asp?
        inc=100100/10.01.00FastForward1&issue=10.01.00
      • Justin Chapweske
        I m probably beating a dead horse here, but there are alot of new people on this list. Going back to a paper I wrote a couple of months ago about P2P
        Message 3 of 17 , Oct 16, 2000
          I'm probably beating a dead horse here, but there are alot of new people
          on this list. Going back to a paper I wrote a couple of months ago about
          P2P interopability, one of my firm suggestions is that that P2P
          applications should stay well away from URLs for linking. Any address
          within the cloud should be independant of specific machines because of the
          dynamic nature of the system. Instead I would like to see a set of URN
          naming schemes proposed and standardized upon so that information can flow
          between otherwise incompatible applications.

          Perhaps this brings up a worthwhile topic of conversation:

          What pieces of Web architecture besides URLs do not meet the needs of P2P
          applications?

          -Justin

          On Tue, 17 Oct 2000, Adam Rifkin wrote:

          > > Having an interoperability layer (analogous to the kinds of
          > > de-facto standards that kicked off the web revolution)
          > > would make things much more interesting.
          >
          > What do you think of the idea of using The Web as that
          > interoperability layer? Run The Web in both directions -- what Tim
          > Berners-Lee and Dave Winer have called the "Two-Way Web" -- and we
          > get standardization *and* interoperability in one swell foop.
          >
          > The world now seems ready for peer-to-peer HTTP networking. Jon
          > Udell was ahead of his time...
          >
          > http://www.byte.com/column/BYT20000830S0001
          >
          > ----
          > Adam@...
          >
          > There are no successful fully distributed applications out there
          > except for Gnutella - in that, it's really a revolution.
          > -- Gene Kan, http://www.internetworld.com/article_bot.asp?
          > inc=100100/10.01.00FastForward1&issue=10.01.00
          >
          >
          >
          > To unsubscribe from this group, send an email to:
          > decentralization-unsubscribe@egroups.com
          >
          >
          >
        • Todd Boyle
          ... The browser. Remember the browser wars in which everybody realized, my GOD, we can write applications to the BROWSER instead of Win32? Microsoft won.
          Message 4 of 17 , Oct 16, 2000
            > What pieces of Web architecture besides URLs do not meet the
            > needs of P2P applications?

            The browser. Remember the browser wars in which everybody
            realized, my GOD, we can write applications to the BROWSER
            instead of Win32?

            Microsoft won. Netscape did too. They all agreed the browser
            should be a dumb client for top down selling, and total control
            by servers. No persistent data, no decent set of widgets for
            writing applications, no keyboard operation for data entry,
            inadequate security infrastructure, etc.

            -TOdd
          • Dave Winer
            Todd, it s even worse than it appears. Your data is locked up on the server. Just like the old mainframe days.
            Message 5 of 17 , Oct 16, 2000
              Todd, it's even worse than it appears.

              Your data is locked up on the server.

              Just like the old mainframe days.

              http://radiodiscuss.userland.com/mainframesAreComputersToo

              Dave
            • Dave Winer
              OK, is it time to get concrete? Here s a scenario.. I have users who create content on their desktop machines. They have a simple text editor with an
              Message 6 of 17 , Oct 16, 2000
                OK, is it time to get concrete?

                Here's a scenario..

                I have users who create content on their desktop machines.

                They have a simple text editor with an integrated web server.

                When they save documents in the www folder, those documents are available
                over HTTP.

                They choose Disconnect from the Web Server menu.

                What happens?

                Note, their content of course must remain accessible with the same address
                as it had when they were connected.

                Dave



                ----- Original Message -----
                From: "Rahul Dave" <rahul@...>
                To: <decentralization@egroups.com>
                Sent: Monday, October 16, 2000 5:56 PM
                Subject: Re: [decentralization] Re: Standardization vs. Interoperability


                > I got this from you:
                > >
                > > I'm probably beating a dead horse here, but there are alot of new people
                > > on this list. Going back to a paper I wrote a couple of months ago
                about
                > > P2P interopability, one of my firm suggestions is that that P2P
                > > applications should stay well away from URLs for linking. Any address
                > > within the cloud should be independant of specific machines because of
                the
                > > dynamic nature of the system. Instead I would like to see a set of URN
                > > naming schemes proposed and standardized upon so that information can
                flow
                > > between otherwise incompatible applications.
                > This would depend upon your namespacing..xns.org's scheme is one example..
                > >
                > > Perhaps this brings up a worthwhile topic of conversation:
                > >
                > > What pieces of Web architecture besides URLs do not meet the needs of
                P2P
                > > applications?
                > DNS, in its direct form..P2P will need, at the least dynamic DNS, and
                better,
                > non global naming. But this is related to namespacing above.(Note that I
                > am not saying get rid of DNS, but that it ought to be hidden from the
                application layer)
                > Rahul
                >
                > To unsubscribe from this group, send an email to:
                > decentralization-unsubscribe@egroups.com
                >
                >
                >
              • Rahul Dave
                ... This would depend upon your namespacing..xns.org s scheme is one example.. ... DNS, in its direct form..P2P will need, at the least dynamic DNS, and
                Message 7 of 17 , Oct 16, 2000
                  I got this from you:
                  >
                  > I'm probably beating a dead horse here, but there are alot of new people
                  > on this list. Going back to a paper I wrote a couple of months ago about
                  > P2P interopability, one of my firm suggestions is that that P2P
                  > applications should stay well away from URLs for linking. Any address
                  > within the cloud should be independant of specific machines because of the
                  > dynamic nature of the system. Instead I would like to see a set of URN
                  > naming schemes proposed and standardized upon so that information can flow
                  > between otherwise incompatible applications.
                  This would depend upon your namespacing..xns.org's scheme is one example..
                  >
                  > Perhaps this brings up a worthwhile topic of conversation:
                  >
                  > What pieces of Web architecture besides URLs do not meet the needs of P2P
                  > applications?
                  DNS, in its direct form..P2P will need, at the least dynamic DNS, and better,
                  non global naming. But this is related to namespacing above.(Note that I
                  am not saying get rid of DNS, but that it ought to be hidden from the application layer)
                  Rahul
                • Rahul Dave
                  ... I wouldnt want the full widgetry either, but rather a compromise between a y2k browser and a windows app. I want users to be able to use megawidgets like
                  Message 8 of 17 , Oct 16, 2000
                    I got this from you:
                    >
                    > > What pieces of Web architecture besides URLs do not meet the
                    > > needs of P2P applications?
                    >
                    > The browser. Remember the browser wars in which everybody
                    > realized, my GOD, we can write applications to the BROWSER
                    > instead of Win32?
                    >
                    > Microsoft won. Netscape did too. They all agreed the browser
                    > should be a dumb client for top down selling, and total control
                    > by servers. No persistent data, no decent set of widgets for
                    > writing applications, no keyboard operation for data entry,
                    > inadequate security infrastructure, etc.
                    I wouldnt want the full widgetry either, but rather a compromise
                    between a y2k browser and a windows app. I want users to be able to
                    use megawidgets like calendars and visiting cards
                    just like they use html tags, and script
                    them with no more than javascript or another scripting language.

                    Thats why I like XUL and XBL in mozilla...
                    Rahul
                    >
                    > -TOdd
                    >
                    > To unsubscribe from this group, send an email to:
                    > decentralization-unsubscribe@egroups.com
                    >
                    >
                    >
                  • Justin Chapweske
                    I partially agree with you. Dynamic DNS is needed extend the utility of URL based naming. But DNS as a system that purely maps names to IP addresses doesn t
                    Message 9 of 17 , Oct 16, 2000
                      I partially agree with you. Dynamic DNS is needed extend the utility of
                      URL based naming. But DNS as a system that purely maps names to IP
                      addresses doesn't go far enough. In systems like Freenet, you need
                      mappings to content hashes, not IP addresses. And there are many who
                      doubt that the DNS can scale to the more fine-grained updates that
                      dynamic DNS demands. I'm of the opinion that it is the P2P applications
                      burdon to scale its own naming system, not the poor sys-admins that
                      already have enough ulcers.

                      -Justin

                      > DNS, in its direct form..P2P will need, at the least dynamic DNS, and better,
                      > non global naming. But this is related to namespacing above.(Note that I
                      > am not saying get rid of DNS, but that it ought to be hidden from the application layer)
                      > Rahul
                      >
                      > To unsubscribe from this group, send an email to:
                      > decentralization-unsubscribe@egroups.com
                      >
                      >
                      >
                    • Rahul Dave
                      Ok, lets tackle this scenario. Suppose I wrote an essay on linking, and called it Linking when I saved it into my www folder. Now saving can be write back or
                      Message 10 of 17 , Oct 16, 2000
                        Ok,
                        lets tackle this scenario.

                        Suppose I wrote an essay on linking, and called it "Linking" when I saved
                        it into my www folder.

                        Now saving can be write back or write through, in the latter case it gets
                        written to the cloud. Or in the former I save to cloud at disconnection.

                        So suppose I had a namespace prefix called rahul, then this document
                        ought to be accessible as rahul.Linking. In my own app, the default
                        namespace is rahul, but for others its not.

                        Now the question is, how is the namespace rahul resolved. To answer that,
                        lets ask, who asks for resolution?

                        The resolution can be done either by a resolving proxy website as done in XNS.org
                        or by a local resolver built into the app. This resolver would first try
                        to contact my local machine, and when that fails as I disconnected, connect
                        to my "virtual presence" in the cloud. In the simple case, this is a
                        known machine in the cloud, stored at my ISP or some website.
                        On contact of that machine it would look for all files in the namespace(folder)
                        rahul(or use some metadata indexing) and get Linking.

                        In the more complex case of a freenet like scenario where there is no well
                        known host of files, and they are all spread out the app contacts the nearest
                        peers asking the question: do you have any files in the namespace(folder)
                        rahul, and is any named linking, (or do you know any other peers who know
                        anything about this rahul fellow, give me their ip's...).

                        Now we have to guarantee that this was really the real rahul, so rahul
                        computes as a part of save, a digital signature on rahul, and on saving
                        or disconnecting, the metadata with this signature and other file(object)
                        parameters is also published to the cloud. Now if a peer or app wants to
                        check the authenticity of the Linking document, they compute its hash
                        and verify the signature using Rahul's public key. And how did they get that?
                        On their first visit to Rahul's app or virtual presence...
                        Rahul
                        I got this from you:
                        >
                        > OK, is it time to get concrete?
                        >
                        > Here's a scenario..
                        >
                        > I have users who create content on their desktop machines.
                        >
                        > They have a simple text editor with an integrated web server.
                        >
                        > When they save documents in the www folder, those documents are available
                        > over HTTP.
                        >
                        > They choose Disconnect from the Web Server menu.
                        >
                        > What happens?
                        >
                        > Note, their content of course must remain accessible with the same address
                        > as it had when they were connected.
                        >
                        > Dave
                        >
                        >
                        >
                        > ----- Original Message -----
                        > From: "Rahul Dave" <rahul@...>
                        > To: <decentralization@egroups.com>
                        > Sent: Monday, October 16, 2000 5:56 PM
                        > Subject: Re: [decentralization] Re: Standardization vs. Interoperability
                        >
                        >
                        > > I got this from you:
                        > > >
                        > > > I'm probably beating a dead horse here, but there are alot of new people
                        > > > on this list. Going back to a paper I wrote a couple of months ago
                        > about
                        > > > P2P interopability, one of my firm suggestions is that that P2P
                        > > > applications should stay well away from URLs for linking. Any address
                        > > > within the cloud should be independant of specific machines because of
                        > the
                        > > > dynamic nature of the system. Instead I would like to see a set of URN
                        > > > naming schemes proposed and standardized upon so that information can
                        > flow
                        > > > between otherwise incompatible applications.
                        > > This would depend upon your namespacing..xns.org's scheme is one example..
                        > > >
                        > > > Perhaps this brings up a worthwhile topic of conversation:
                        > > >
                        > > > What pieces of Web architecture besides URLs do not meet the needs of
                        > P2P
                        > > > applications?
                        > > DNS, in its direct form..P2P will need, at the least dynamic DNS, and
                        > better,
                        > > non global naming. But this is related to namespacing above.(Note that I
                        > > am not saying get rid of DNS, but that it ought to be hidden from the
                        > application layer)
                        > > Rahul
                        > >
                        > > To unsubscribe from this group, send an email to:
                        > > decentralization-unsubscribe@egroups.com
                        > >
                        > >
                        > >
                        >
                        >
                        > To unsubscribe from this group, send an email to:
                        > decentralization-unsubscribe@egroups.com
                        >
                        >
                        >
                      • Wesley M. Felter
                        ... This scenario could use a reverse Web proxy cache somewhere upstream on the network. If the user is connected, the proxy can just pass requests through to
                        Message 11 of 17 , Oct 16, 2000
                          On Mon, 16 Oct 2000, Dave Winer wrote:

                          > Here's a scenario..
                          >
                          > I have users who create content on their desktop machines.
                          >
                          > They have a simple text editor with an integrated web server.
                          >
                          > When they save documents in the www folder, those documents are available
                          > over HTTP.
                          >
                          > They choose Disconnect from the Web Server menu.
                          >
                          > What happens?
                          >
                          > Note, their content of course must remain accessible with the same address
                          > as it had when they were connected.

                          This scenario could use a reverse Web proxy cache somewhere upstream on
                          the network. If the user is connected, the proxy can just pass requests
                          through to the user's desktop Web server. If the users is disconnected,
                          the proxy can serve stuff out of its cache. This also helps if the user
                          gets slashdotted.

                          If you want to be even more clever, combine this with Rahul's idea to have
                          a smart DNS server that points to the user's machine when it is connected
                          but the cache when the user is disconnected.

                          This isn't a very decentralized solution IMO, but that's what you get when
                          you need interoperability with the Web.

                          Wesley Felter - wesf@... - http://www.cs.utexas.edu/users/wesf/
                        • Rahul Dave
                          Oh, I dont think dynamic DNS is the answer; it is the minimally needed thing..Manila does this or something similar for editthispage, is there something
                          Message 12 of 17 , Oct 16, 2000
                            Oh, I dont think dynamic DNS is the answer; it is the minimally needed
                            thing..Manila does this or something similar for editthispage, is there
                            something similar where rahul.radiouserland.com maps to my machine(and
                            falls back to some kinda presence at Userland)?
                            I got this from you:
                            >
                            > I partially agree with you. Dynamic DNS is needed extend the utility of
                            > URL based naming. But DNS as a system that purely maps names to IP
                            > addresses doesn't go far enough. In systems like Freenet, you need
                            > mappings to content hashes, not IP addresses. And there are many who
                            I'm utterly unconvinced about using content hashes for naming, but I'd
                            rather use then for verification of content and metadata..
                            > doubt that the DNS can scale to the more fine-grained updates that
                            > dynamic DNS demands. I'm of the opinion that it is the P2P applications
                            > burdon to scale its own naming system, not the poor sys-admins that
                            I agree.

                            XNS.org provides an interesting global scheme.

                            The angle I am pursuing is local, where each peer defines local names;
                            basically I want to try mixing SDSI and capabilities..

                            SDSI is public key crypto plus local naming where a name is mapped to a
                            public key. Capabilities define statements like "the person who in
                            my local database maps to dave may read 'Linking'"
                            > already have enough ulcers.
                            >
                            > -Justin
                            >
                            > > DNS, in its direct form..P2P will need, at the least dynamic DNS, and better,
                            > > non global naming. But this is related to namespacing above.(Note that I
                            > > am not saying get rid of DNS, but that it ought to be hidden from the application layer)
                            > > Rahul
                            > >
                            > > To unsubscribe from this group, send an email to:
                            > > decentralization-unsubscribe@egroups.com
                            > >
                            > >
                            > >
                            >
                            >
                            > To unsubscribe from this group, send an email to:
                            > decentralization-unsubscribe@egroups.com
                            >
                            >
                            >
                          • Clay Shirky
                            ... The URL scheme is far more flexible than DNS. URLs could easily be extended to the form napster://green_day_fan or icq://12345678. ... This is already
                            Message 13 of 17 , Oct 16, 2000
                              > I partially agree with you. Dynamic DNS is needed extend the utility of
                              > URL based naming.

                              The URL scheme is far more flexible than DNS. URLs could easily be
                              extended to the form napster://green_day_fan or icq://12345678.

                              > But DNS as a system that purely maps names to IP addresses doesn't
                              > go far enough. In systems like Freenet, you need mappings to
                              > content hashes, not IP addresses. And there are many who doubt that
                              > the DNS can scale to the more fine-grained updates that dynamic DNS
                              > demands. I'm of the opinion that it is the P2P applications burdon
                              > to scale its own naming system, not the poor sys-admins that already
                              > have enough ulcers.

                              This is already happening. Napster et al have moved from
                              machine-centric naming schemes to protocol-centric addressing schemes,
                              and there are now more protocol-centric addresses in use than gTLD
                              addresses. (Order 150 million compared to order 15 million).

                              http://www.shirky.com//writings/dns.html

                              --
                              Clay Shirky | shirky.com - Essays on the Internet:
                              http://www.shirky.com/ | Culture, Economics, Globalization
                            • Justin Chapweske
                              Read this post twice, because I think it clarifies alot of issues: ... I should change my original statement to be Dynamic DNS is needed to extend the utility
                              Message 14 of 17 , Oct 16, 2000
                                Read this post twice, because I think it clarifies alot of issues:

                                On Mon, 16 Oct 2000, Clay Shirky wrote:

                                > > I partially agree with you. Dynamic DNS is needed extend the utility of
                                > > URL based naming.
                                >
                                > The URL scheme is far more flexible than DNS. URLs could easily be
                                > extended to the form napster://green_day_fan or icq://12345678.
                                >

                                I should change my original statement to be "Dynamic DNS is needed to
                                extend the utility of *interopable* URL based naming." The problem with
                                the examples you provide is that they are protocol-specific naming
                                mechanisms and therefore not interopable w/o building an uber-browser that
                                implements every silly protocol on the planet. I'm glad that you used
                                this example because this is exactly why I think URLs should NOT be used
                                in P2P systems.

                                Unlike URLs which define protocol schemes, URNs define *naming
                                schemes*. While the number of useful protocols is unlimited, the set of
                                useful naming schemes is actually rather small. And having a limited set
                                of schemes is imperative for interopability.

                                So for URN schemes we might have a query scheme, a content-hash scheme,
                                and Rahul's local namespace scheme. A next generation browser would then
                                implement these different schemes, and because of the flexibility allowed
                                in their implementation I think we would see some REALLY interesting
                                solutions for dealing with the name scaling problems.

                                >
                                > This is already happening. Napster et al have moved from
                                > machine-centric naming schemes to protocol-centric addressing schemes,
                                > and there are now more protocol-centric addresses in use than gTLD
                                > addresses. (Order 150 million compared to order 15 million).
                                >

                                That is a very intersting point, definately something to chew on. But my
                                question is how the heck am I supposed send a link to an ICQ message that
                                points to an mp3 sitting in Napster? <emphasis> My point is that
                                addresses by themselves are useless, its the links that are
                                important. There are certainly many times more useful links between gTLD
                                addresses than protocol specific addresses. </emphasis>

                                The value of a system is measured by the number of interconnections
                                between components, not the raw number of components.

                                -Justin
                              • Clay Shirky
                                ... As they say at the Media Lab Anything you can do, I can do meta. No need for an uber-browser, just a standards based version of what RealNames does. See
                                Message 15 of 17 , Oct 17, 2000
                                  > The problem with the examples you provide is that they are
                                  > protocol-specific naming mechanisms and therefore not interopable
                                  > w/o building an uber-browser...

                                  As they say at the Media Lab "Anything you can do, I can do meta."

                                  No need for an uber-browser, just a standards based version of what
                                  RealNames does. "See something in the command line that isn't a URL?
                                  Send it to this address and we'll resolve it remotely."

                                  That could be extended to protocols.

                                  > That is a very intersting point, definately something to chew on. But my
                                  > question is how the heck am I supposed send a link to an ICQ message that
                                  > points to an mp3 sitting in Napster?

                                  aimster.com is working on precisely this problem. Take a look.

                                  > The value of a system is measured by the number of interconnections
                                  > between components, not the raw number of components.

                                  Agreed, but I think the issue is already settled. Developers are
                                  creating protocol-specific namespaces by the bushel. Arguing that they
                                  _shouldn't_ do such a thing is like re-capping the argument that HTML
                                  shouldn't contain descripting tags, since that violates SGML's
                                  semantics-only dictum.

                                  URNs are a sensible, logical, perfectly adaptable non-starter of an
                                  idea. The problem here, as with so much on the net, is that the
                                  engineering revolution moves faster than the theoretical framework, so
                                  by the time we know just how something should be implemented, its too
                                  late.

                                  --
                                  Clay Shirky | shirky.com - Essays on the Internet:
                                  http://www.shirky.com/ | Culture, Economics, Globalization
                                • Rahul Dave
                                  ... See www.xns.org. The idea is to use a local resolving agent or a virtual resolving agent in the cloud to resolve a name to a protocol.. ... How are urn s
                                  Message 16 of 17 , Oct 17, 2000
                                    I got this from you:
                                    >
                                    > > The problem with the examples you provide is that they are
                                    > > protocol-specific naming mechanisms and therefore not interopable
                                    > > w/o building an uber-browser...
                                    >
                                    > As they say at the Media Lab "Anything you can do, I can do meta."
                                    >
                                    > No need for an uber-browser, just a standards based version of what
                                    > RealNames does. "See something in the command line that isn't a URL?
                                    > Send it to this address and we'll resolve it remotely."
                                    See www.xns.org. The idea is to use a local resolving agent or a virtual resolving agent
                                    in the cloud to resolve a name to a protocol..
                                    >
                                    > That could be extended to protocols.
                                    >
                                    > > That is a very intersting point, definately something to chew on. But my
                                    > > question is how the heck am I supposed send a link to an ICQ message that
                                    > > points to an mp3 sitting in Napster?
                                    >
                                    > aimster.com is working on precisely this problem. Take a look.
                                    >
                                    > URNs are a sensible, logical, perfectly adaptable non-starter of an
                                    > idea. The problem here, as with so much on the net, is that the
                                    > engineering revolution moves faster than the theoretical framework, so
                                    > by the time we know just how something should be implemented, its too
                                    > late.
                                    How are urn's so different from real names, or xns names, or even locally generated names
                                    such as in SDSI...SDSI uses hash of public key, xns a GUID, and real names I dont know(and
                                    dont care since their revenuemodel is selling names)? All URN's do is point to a resource,
                                    and they are not linked to http to the best of my knowledge..a URN can basically be a GUID..
                                    or even a java package scoped name.

                                    In my estimation, for interoperability we must agree on a naming convention: any of these
                                    can suffice for the other. The issues we must consider, I think are
                                    (a) should naming be global or local(and even here local names can be mapped to
                                    global ones?
                                    (b) who should control naming(not real names, not ICANN, not verisign, plays back into (a))
                                    This is as much a political question as a technological one, and rife with trademark type
                                    issues, which is why I like to push a local system..if i want to call ibm.data(as according
                                    to verisign) bigblue.data, its my concern, and when things are downloaded by others, it is
                                    seen that I call bmi what verisign calls ibm, and because the downloader may decide to trust
                                    verisign's judgement in certifying ibm's public key there, the right data is obtained..

                                    Rahul
                                  • TurboJerry
                                    ... An IRC channel + logs would be useful. Jeremy
                                    Message 17 of 17 , Nov 26, 2000
                                      --- In decentralization@egroups.com, "Tim O'Reilly" <tim@o...> wrote:
                                      > b) What we really need is just a forum for people to get together to
                                      > start conversations about interoperability.

                                      An IRC channel + logs would be useful.

                                      Jeremy
                                    Your message has been successfully submitted and would be delivered to recipients shortly.