/[gzz]/manuscripts/URN5/urn5.rst
ViewVC logotype

Diff of /manuscripts/URN5/urn5.rst

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 1.1.1.1 by benja, Fri Dec 6 15:56:57 2002 UTC revision 1.2 by tjl, Thu Dec 12 18:42:01 2002 UTC
# Line 11  ABSTRACT Line 11  ABSTRACT
11  ========  ========
12    
13  Our research group has recently registered the informal ``urn-5`` namespace  Our research group has recently registered the informal ``urn-5`` namespace
14  with IANA. We discuss the rationale and  with IANA.
15    We discuss the rationale and
16  possible applications for this open namespace.  possible applications for this open namespace.
17    
18    The namespace is basically random numbers, important is the semantics
19    and applications.
20    
21    
22  Introduction  Introduction
23  ============  ============
# Line 38  Benja says: Line 42  Benja says:
42  - URN  - URN
43  - OID!!  - OID!!
44    
45    Ontologies
46    
47  Needs for identification  Needs for identification
48      RDF      RDF
49      bidirectional linking      bidirectional linking
50    
51    Content-addressability and pointing: the two complementary operations
52    =====================================================================
53    
54    "I want the newspaper article whose hash is 239273498275"
55    "I want to know what articles are in today's newspaper"
56    
57    Normal WWW conflates the two.
58    E.g. WWW caching: cache will ask original server whether something
59    changed and if not, bring the cached document.
60    If two operations separated, more flexible mechanism: always
61    ask original server for name --> content-address mapping, then
62    ask P2P network for actual data.
63    
64    You mean, talking about the idea that all user-meaningful "objects"
65    (Ted's items...), like documents, paragraphs, vector graphics shapes
66    etc. get a global identity the moment they're created, allowing to talk
67    about them in a globally unique way? That sounds interesting. Could
68    relate this to the Semantic Web (how you're able to make RDF statements
69    about these things, e.g. metadata). Talk about how it gets rid of
70    context problems (the id stays the same when moving the object to a
71    different document) and possibly, how this allows for Ted's
72    transclusions and links (links can be implemented using RDF and RDF
73    search engines, which are in development). Can then relate it to
74    zzstructure (editing textual XML with these identifiers is possible, but
75    no fun; with a structural editor, like a zzstructure client, the ids can
76    remain hidden).
77    
78    Yes. Along with discussing the difference between this and urn:sha-1 really
79    carefully.
80    
81    
82    (Between this and "identification schemes based on secure hashing, such
83    as the (unregistered) Freenet URIs": urn:sha-1 is neither used nor
84    registered nor in the process of being registered, so the most we could
85    refer to it is as an idea someone has had at some point. A nit, I know,
86    but important when publishing this.)
87    
88    Possible order of presentation:
89    
90    - Globally unique ids and why you want to use them
91    - Places that don't use globally unique ids currently, and what you
92      cannot do because of that
93    
94    
95    Cite some of the DNS troubles, muscling of individuals by large corporations
96    etc.
97    
98  URN namespaces  URN namespaces
99  ==============  ==============
100    
101  ALL OTHER CURRENT URNs ARE CLOSED NAMESPACES?!?!  ALL OTHER CURRENT URNs ARE CLOSED NAMESPACES?!?!
102  http://www.iana.org/assignments/urn-namespaces  http://www.iana.org/assignments/urn-namespaces
103  EXCEPT urn:sha WHICH IS NOT OFFICIAL  EXCEPT urn:sha WHICH IS NOT OFFICIAL, and is a content-addressing name.
104    
105  If you have a domain name, you can use urn:NewsML:, and the  If you have a domain name, you can use urn:NewsML:, and the
106  urn:tag: currently submitted to IESG for publication as an RFC.  urn:tag: currently submitted to IESG for publication as an RFC.
# Line 77  Should we go through all the urn types b Line 130  Should we go through all the urn types b
130    
131    
132    
133    
134  The ``urn:urn-5:`` namespace  The ``urn:urn-5:`` namespace
135  ============================  ============================
136    
137  CONTENT-ADDRESSABILITY AND POINTING: THE TWO COMPLEMENTARY OPERATIONS  - The urn:urn-5 namespace
 "I want the newspaper article whose hash is 239273498275"  
 "I want to know what articles are in today's newspaper"  
138    
139  You mean, talking about the idea that all user-meaningful "objects"  - how it works
 (Ted's items...), like documents, paragraphs, vector graphics shapes  
 etc. get a global identity the moment they're created, allowing to talk  
 about them in a globally unique way? That sounds interesting. Could  
 relate this to the Semantic Web (how you're able to make RDF statements  
 about these things, e.g. metadata). Talk about how it gets rid of  
 context problems (the id stays the same when moving the object to a  
 different document) and possibly, how this allows for Ted's  
 transclusions and links (links can be implemented using RDF and RDF  
 search engines, which are in development). Can then relate it to  
 zzstructure (editing textual XML with these identifiers is possible, but  
 no fun; with a structural editor, like a zzstructure client, the ids can  
 remain hidden).  
140    
141    The "technology" behind the ``urn-5`` namespace is more or less trivial:
142    the identifier structure is ::
143    
144  One really important point:           urn:urn-5:<random number>[:<local part>]
145    
146      non-hierarchy.  where *<random number>* is a random number of at least 160 bits
147    base64-encoded. The local part is an arbitrary string of characters
148    allowed in URNs without escaping.
149    
 Cite some of the DNS troubles, muscling of individuals by large corporations  
 etc.  
150    
151    Creating globally unique "nodes".
152    
153    Any time, anywhere, no hassle whatsoever, no configuration, *NOTHING*.
154    
155  Yes. Along with discussing the difference between this and urn:sha-1 really  One really important point:
 carefully.  
156    
157        non-hierarchy.
158    
159    Unguessable
160        
161        - except when using ...-1, ...-2
162            
163            - should only do so when certain it's not a problem.
164    
165                - ZZ spaces: ??!
166    
167    
 (Between this and "identification schemes based on secure hashing, such  
 as the (unregistered) Freenet URIs": urn:sha-1 is neither used nor  
 registered nor in the process of being registered, so the most we could  
 refer to it is as an idea someone has had at some point. A nit, I know,  
 but important when publishing this.)  
168    
 Possible order of presentation:  
169    
 - Globally unique ids and why you want to use them  
 - Places that don't use globally unique ids currently, and what you  
   cannot do because of that  
 - The urn:urn-5 namespace  
 - how it works  
 - how it's different from hashing  
170  - Caveat: Human readability/writability problems  - Caveat: Human readability/writability problems
 - only in text editors, structure editors can handle ids internally  
171    
172  - Caveat: unlike relative identifiers, when you have an urn:urn-5 id,      - only in text editors, structure editors can handle ids internally
173    you don't automatically have any document/context it appears in  
174    - when an anchor is removed?  Anyone may "spoof" an id -- but not really mean that because
175    urn-5 SHOULD NOT BE USED FOR DIRECT RESOLUTION WITH ONE RESULT!!
176    
177    Possibly easier to think of as "words": using urn-5 names for concepts
178    makes it easier for search engines. For instance, searching for web pages
179    discussing the ancient board game "go" is made difficult by the common
180    homonym.
181    
182    If the word had been agreed to be ``urn:urn-5:Ot877naD3qzF5VzzItZbozeKWOuV'',
183    no such problem would arise. Of course, an intentional DOS attack
184    of creating many pages seemingly discussing this would be possible.
185    
186    The URN-5 name could be seen as a node in a graph, and resolving the name
187    would be asking for the edges it touches.
188        
189        urn-5s,
190        by themselves, *never* refer to documents.
191        They are not resolvable. Documents may only state that they have
192        something to do with a urn-5, and a search may be performed and return
193        that; but that doesn't mean that you can make a reference,
194        'show the document identified by this urn-5, here.'
195    
 Cryptography  
 ============  
196    
 Tying a particular URN-5 to an identity using signatures?  
197    
198  Applications  Applications
199  ============  ============
# Line 147  Applications Line 202  Applications
202    you want to perform lookups using them    you want to perform lookups using them
203  - this kind of registry would allow RDF statements about identified  - this kind of registry would allow RDF statements about identified
204    objects to be resolved    objects to be resolved
205    
206  - could use this to implement transclusions and links  - could use this to implement transclusions and links
207  - Conclusions      - NOT!
208    
209  - RDF  - RDF
210  - HTML anchors  - HTML anchors
# Line 161  used in places where urn-5 could be used Line 217  used in places where urn-5 could be used
217    
218  (Btw, JXTA ids may also be worth looking at.)  (Btw, JXTA ids may also be worth looking at.)
219    
220    - Caveat: unlike relative identifiers, when you have an urn:urn-5 id,
221      you don't automatically have any document/context it appears in
222      - when an anchor is removed?
223    
224    Identity: libpaper
225    
226  Potential problems  Conclusion
227  ==================  ==========
228    
229  Potential problem: someone makes program that assigns urn-5s badly,  Potential problem: someone makes program that assigns urn-5s badly,
230  to create havoc. This problem is really no different from the corresponding  to create havoc. This problem is really no different from the corresponding
231  one with digital signatures: you have to trust the software you use.  one with digital signatures: you have to trust the software you use.
232    
233  Tuomas once said:  However, the worst result is that some of the nodes in the graph
234      Another: no way to verify an index. Anyone may spoof an id easily.  described above are joined. While annoying, it should be no catastrophen
     Therefore, you can't really refer to an id outside the current context:  
     too easy to substitute another document for it.  
   
 Benja notes:  
     As was clarified in the following discussion, Tuomas' statement above  
     *makes no sense*. urn-5s, by themselves, *never* refer to documents.  
     They are not resolvable. Documents may only state that they have  
     something to do with a urn-5, and a search may be performed and return  
     that; but that doesn't mean that you can make a reference,  
     'show the document identified by this urn-5, here.'  
235    
     We have to be really careful to get this point across correctly,  
     beca  
   
 T:  
   
     Someone can still "DOS" attack a URN-5 by creating lots of documents  
     that talk about it as something entirely different from the original.  
       
236    
237  Coupling urn-5 with digital signatures?  Coupling urn-5 with digital signatures?
238    
# Line 211  Benja says: Line 253  Benja says:
253      (Freenet gets around the problem by identifying entities      (Freenet gets around the problem by identifying entities
254      by unchangeable public keys; not nice.)      by unchangeable public keys; not nice.)
255    
256  When central, X.400, PKI  When central, X.400, PKI.
257    However, then why use URN-5 at all if you have central names?
258    
   
 Conclusion  
 ==========  

Legend:
Removed from v.1.1.1.1  
changed lines
  Added in v.1.2

savannah-hackers-public@gnu.org
ViewVC Help
Powered by ViewVC 1.1.26