/[gzz]/manuscripts/pointers/article.rst
ViewVC logotype

Diff of /manuscripts/pointers/article.rst

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

revision 1.139 by benja, Thu Nov 6 07:01:35 2003 UTC revision 1.140 by benja, Thu Nov 6 07:11:51 2003 UTC
# Line 208  The main contribution of this paper is t Line 208  The main contribution of this paper is t
208  a versioning mechanism which is similar to OceanStore's heartbeats,  a versioning mechanism which is similar to OceanStore's heartbeats,
209  but allows the clients to download and store the pointer records  but allows the clients to download and store the pointer records
210  along with the corresponding version of a document.  along with the corresponding version of a document.
 If a client caches a version of a document locally, it  
 stores the pointer record along with it.  
211  Instead of having a primary replica, we search  Instead of having a primary replica, we search
212  the P2P network for current pointer records  the whole P2P network for pointer records
213  related to a document; we pick the one with  related to a document.
 the most current timestamp as the version  
 to show to the user by default; old versions  
 remain equally accessible  
 as long as  
 anybody keeps a copy.  
214    
215  An additional contribution is the Storm data model,  An additional contribution is the Storm data model,
216  an API formalizing the notion of hash-based addressing  an API formalizing the notion of searching for data
217  and content-based lookup that can be implemented  by hash and content. The API is used by applications
218  using many different P2P architectures (including  such as browsers and implemented for each P2P network
219  DHTs and the flooding broadcast search of Gnutella [gnutellaurl]_).  accessed. Pointer records can be implemented
220  The Storm data model provides a convenient interface  on top of the Storm model, rather than separately
221  to applications such as Web browsers, which then  for each P2P network.
 need not worry about the P2P network(s) used to  
 resolve the identifiers. Pointer records can be implemented  
 on top of the Storm model, meaning that they do not  
 need to be implemented separately for each time of  
 P2P network.  
222    
223  ..  Like in OceanStore, old  ..  Like in OceanStore, old
224      versions remain fully accessible (as long as somebody      versions remain fully accessible (as long as somebody
# Line 243  generated through server-side scripting) Line 231  generated through server-side scripting)
231  would have the benefits of semantic-free referencing--  would have the benefits of semantic-free referencing--
232  URIs don't have to change because of trademark conflicts  URIs don't have to change because of trademark conflicts
233  as in DNS, for example-- but go beyond the proposal  as in DNS, for example-- but go beyond the proposal
234  by Balakrishnan et.al. [balakrishnan03semanticfree]_,  by Balakrishnan et.al. [balakrishnan03semanticfree]_
235  for replacing DNS but keeping the centralized architecture of  to replace DNS but keep the centralized architecture of
236  one Web server (or a single set of servers) maintained  one Web server (or a single set of servers) maintained
237  by the publisher of a document.  by the publisher of a document.
238    
# Line 324  Because they require no centralized vers Line 312  Because they require no centralized vers
312  pointer records may be able to unify the namespaces  pointer records may be able to unify the namespaces
313  of different P2P systems in a similar way to hash-based identifiers.  of different P2P systems in a similar way to hash-based identifiers.
314  Similar to searching different P2P networks in order  Similar to searching different P2P networks in order
315  to resolve a SHA-1 hash, you can search different networks  to resolve a SHA-1 hash, you could search different networks
316  to find pointer records for a document. A document  to find pointer records for a document. A document
317  could be originally edited in OceanStore, then published  could be originally edited in OceanStore, then published
318  in Gnutella or a DHT-based system, using an anonymized system  in Gnutella or a DHT-based system, using an anonymized system
# Line 395  Reverse indices Line 383  Reverse indices
383      The server would publish the resulting mappings      The server would publish the resulting mappings
384      from words to block ids on the P2P network.      from words to block ids on the P2P network.
385    
386  An index can also return additional information  .. XXX kept out because of space constraints:
387  to be placed in the hashtable along with the  
388  block id. In the example, this could be the complete     An index can also return additional information
389  metadata of the song, so that a client does not need     to be placed in the hashtable along with the
390  to download the complete audio file before being able     block id. In the example, this could be the complete
391  to show its artist, title etc. to its user.     metadata of the song, so that a client does not need
392       to download the complete audio file before being able
393       to show its artist, title etc. to its user.
394    
395  The pointer functionality is trivial to implement  The pointer functionality is trivial to implement
396  *on top of* this basic model. A pointer record  *on top of* this basic model. A pointer record

Legend:
Removed from v.1.139  
changed lines
  Added in v.1.140

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