Re: Intelligent Rating Systems

R. Martin Roscheisen (rmr@cs.stanford.edu)
Tue, 27 Jun 1995 11:33:35 -0700


Message-Id: <9506271833.AA28658@Xingu.Stanford.EDU>
To: liberte@ncsa.uiuc.edu (Daniel LaLiberte)
Cc: bede@scotty.mitre.org, rmr@cs.stanford.edu, rating@junction.net,
Subject: Re: Intelligent Rating Systems 
From: "R. Martin Roscheisen" <rmr@cs.stanford.edu>
In-Reply-To: Your message of "Mon, 26 Jun 1995 23:59:14 CDT."
             <9506270459.AA16138@void.ncsa.uiuc.edu> 
Date: Tue, 27 Jun 1995 11:33:35 -0700


  >> You cannot assume that a content provider will be willing to carry
  >> ratings by others.  
  >
  >True, but if they do offer them, why not consider them, with due caution?

Of course, if some content server coincides with a rating/meta server,
then they will be considered in the same way.  The question is how
this should affect the design.  Version A says, let's define two
separate protocols, one to outside rating servers, one to rating
servers coinciding with the document server.  Version B says, let us
define one protocol, and then some implementers might do certain
optimizations by bundling this and that request into one network access.

I personally prefer version B; it seems to be cleaner design, and it
also avoids tying the rating protocol into http more than it is
necessary (Note that we might use CORBA for this protocol with
advantage, or how about a object request protocol to MSN, ...;-).

  >Daniel LaLiberte (liberte@ncsa.uiuc.edu)

Cheers, - RMR