WHOIS++ & centroids
-------------------
(... plus some other stuff :-) 

Martin Hamilton, Loughborough University 

martin@mrrl.lut.ac.uk 


Pros of search and retrieval protocols 

 o can deliver structured query to server. 

 o return of structured results back to client. 

 o facilitates searches across multiple servers. 

 o whilst not typically supported at the moment in WWW browsers, can
   be retro-fitted via proxy servers, plug-ins, or by using (say) CGI
   gateways. 


Cons of search and retrieval protocols 

 o no provision (typically) made for including of presentation
   oriented material in search results - i.e. adverts ! 

 o some search and retrieval protocols are open-ended about query and
   response formats, hindering interoperability. 

 o some query and response formats are very open-ended in themselves,
   which also hinders interoperability. 

 o still very much a specialized interest, rather than a mainstream
   activity like (say) HTTP. 


Where does this leave us ? 

 o Wide variety of protocols to choose from, usually with associated
   query and response formats. 

 o Mainstream interest not necessarily a good match for the sorts of
   things we are trying to do. 

 o Possible to support a variety of protocols and formats, up to a
   point.  Expect some loss of information if deriving everything from
   one database.

 o May be desirable to identify a lowest common denominator (subset?)
   of database elements a la Dublin Core. 


Where next for protocols and formats ? 

 o IAFA template format provided a reasonable starting point for
   simple Internet resource descriptions - low cost of creation,
   easily transcribed into other formats. Dublin Core & Warwick
   Framework the next evolutionary step ? 

 o WHOIS++ protocol attractive as a lowest common denominator search
   and retrieval protocol, and as the driving force behind
   centroids. Still quite young, though - the RFC only came out a year
   ago.

 o History suggests LCDs usually most successful, but other lowest
   common denominator approaches abound, e.g. Harvest and LDAP. Z-Lite
   might be on its way, etc etc. Watch the skies!


What about centroids ? 

 o Provide an abstract characterization of the database in a standard
   format.

 o Can export this to any number of index servers by pushing or
   pulling.

 o Index servers act as brokers between client and potentially
   multiple servers. 

 o Next step on after Harvest Brokers/Gatherers - index servers don't
   need all of the info for a server, just its centroid.

 o Obvious role would be for subject services to act as brokers (index
   servers) for their disciplines. 


Conclusions 

 o Much interest in centroid approach for whole-Web indexing and
   searching.  May want to be able to search index server via multiple
   protocols, and index Web stuff in general.

 o No clear winner in the lowest common denominator search and retrieval
   protocol / data format stakes. More work needed. Heavyweight
   solutions like Z39.50 and X.500 not popular for Internet
   applications due to interoperability problems and (perceived?)
   implementation complexity.

 o Do we even need/want a search and retrieval protocol ? Search and
   retrieval protocol model fundamentally incompatible with
   advertising based revenue model, which is one possible future for
   subject services ?

Martin Hamilton 
Wed Aug  7 00:31:59 BST 1996