[Top] [All Lists]

Request for recommendation on a Content-Describe: tag

1995-05-31 15:46:42

We have been having a discussion (on com-priv(_at_)psi(_dot_)com, 
about 'ratings servers'.  We have already found the word 'ratings' to be
probably defective as we are really talking about generalized classification

We want to define both an HTML tag and a header line, and conceivably a
MIME type.

I'd like to know if anything related to this has been defined and whether
there is any strong opinions on the naming of the tag/type.

Below I've clipped some of my articles that are some unorganized ideas
about possibilities:

Or howbout just <CONTENT Tags>?  For cleanness I suppose that the
tags should be separate, but I feel pretty strongly right now that you
should be able to pack them like this:

<title>Terminator 2</title>
<CONTENT authority=author -L2V1H0S2 Type=movie
  Alt1 SciFi Fx3 Extropian1 Gunfight BodyCount=12/5>

Or maybe:
<CONTENT> authority=author -L2V1H0S2 Type=movie
  Alt1 SciFi Fx3 Extropian1 Gunfight BodyCount=12/5 </CONTENT>
<CONTENT> authority=reviewer/siskel(_at_)someplace_in_chicago
  thumb=up3 content_link=author/agree </CONTENT>

Another example:
<title>Forward Error Correction for Space Probes</title>
<CONTENT authority=librarian Type=book/tech/ref
 LCIS=23.23132 DEWEY=2.33442 Keywords="ecc,Reed-Solomon,XOR,Vitterbi">

Yet another:

Subject: Space probe programming
Content: authority=recruiter Type=job/contract/offer
        industry_code=4435 Location=us/mid-atlantic/dc
        programmer2 mathematician3 security_clearance=US/NASA/confidential

In the last example, you could see how a lot of info could be
defaulted to the newsgroup description (extension of Newsgroups file).

You define a list of concepts in a distributed and unambiguous way.
Allow for multiple CONTENT tags.
Allow for type, degree, value.
Allow for different independent realms of content types.
Try to make things searchable.
Think about how to do comparison/proximity, filtering, overrides or
conflicts, etc.

This is similar to the old Email/News Keywords: header, but I'm
saying we need to go ahead and start defining type realms and
working out details.  There's no reason we can't use the result
elsewhere, as in "Keywords:"...

I know that a lot of this is done in library sciences, and I'm sure
they have some opinions on how to create hierarchies (I hope so,
although as far as I know there is only Dewey and Library of
Congress), but that shouldn't slow us down.

I've been inconsistent on purpose, yea, that's it.

Oops... I just remembered, I believe that MIME uses Content-Type:.
Too close?  CONTENT sounds a lot like it should be the actual
payload, not a description of it.

Content-Describe:, Description:, Concepts:, Content-Summary:,
Classification:, Flavor:, Selectors:, Semantic-Indexes:,
Semantics:, Semantic-Keys:, Meta-Keys:, Information:,
Meta-Info:, Taxonomy:, Exon-Keys:, Exons:, Snoxe:, Slug-Keys:,
Meme-Keys:, Memedex:, Grips:, Cepts:.

Lets choose one or more.  I like several.  Short ones especially.

I've seen something like this locally that used 0-3 for degrees for
each of these.  That way it's more of a descriptor even for adults
than just a censoring flag.  You could certainly elaborate: G for gore
for instance, separate from Violence.

SC Sexual Content
L  ('dirty') Language
A  Adult themes
V  Violence
G  Gore (blood, guts, severed heads/limbs, etc.)
R  Religious themes  (or possibly philosophical)
H  Humor
S  Suspense
Alt Alternative culture


Also, I'd like to see support for using multiple rating servers in
some defined way.  (Fallback, and, or, topic or site range selectors,
etc.)  Change/what's new lists or ratings, support for different types
of protocols (URL's cover almost all of this already though) like news
groups and articles, telnet/port/login combinations, mail
message-id's, etc.

IP ranges aren't covered yet really by URL's so maybe some form of
firewall style filtering formalism cast into a URL might be useful.

Also, it seems that the rating values could take many forms.  A lot of
thought should go into how to extend the typing to suit a number of
uses.  Examples might be: 

suitability for audience (current topic),
legal status by country,
verifiability (ie. fact checkers used?, official source?, printed
  reference?, unretouched photo, certified production level or virus
  free, etc.),
signature key checks,
opinion indexes (movie ratings categories, music style/taste indexes,
  painting style, etc.),
subject indexing (path through some large subject indexes, similar to library
  topic indexing), etc.

Both author provided and subsequent evaluations should be supported.

Ownership/royalty assignment, routing, etc. could also be recorded and
tracked this way, with optional and/or anonymous demographics
feedback.  Need to be very careful with privacy issues of course.

This provides yet another way to pay for the ratings servers.  If you
are going to be hitting a server anyway, what better place to add
the feedback.  If a customer can choose which ratings servers to use
among many, there is a strong incentive to build a good reputation.

This way it could be used for the purpose at hand and also for future
rating/typing/evaluations.  Could easily create a whole new market
for all the librarians, editors, reporters, experts, hobbiests, etc.
grading data into information.  Anything that doesn't allow for this
will just have to be replaced fairly soon.  That doesn't mean of
course that we implement all of this now.  Note that a lot of 
commercial news information is already suitably graded with
categories, slugwords, etc.  (Reuters, UPI)

I think it's very important to be able to use ratings servers both at
the client software level and at intermediate servers and proxies.  A
firewall with proxy servers for instance is a great place to use
ratings servers to filter traffic.  This solves the problem of
migrating client support and users replacing clients.  (Ie. in a
school or ISP situation.)

Another related topic would be support for 'endorsement' types of
entries on ratings servers.  The idea here is that distributed members
contribute 'endorsements', cryptographically signed, to a ratings
network of servers.  You then 'subscribe' to entities that you agree

Stephen D. Williams 25Feb1965 VW,OH (FBI ID) sdw(_at_)lig(_dot_)net
Consultant, Vienna,VA Mar95- 703-918-1491W 43392 Wayside Cir.,Ashburn, VA 22011
OO/Unix/Comm/NN       ICBM/GPS: 39 02 37N, 77 29 16W home, 38 54 04N, 77 15 56W
Pres.: Concinnous Consulting,Inc.;SDW Systems;Local Internet Gateway Co.;28May95

<Prev in Thread] Current Thread [Next in Thread>