Re: about CritLink and annotation systems in general

At 09:41 AM 8/27/1999 -0400, Bjarni R. Einarsson wrote:
>I thought I'd mention an alternate (slightly more hack-ish) way I've
>considered to implement this - which would work with all current browsers
>and allow collaboration with many different services, while avoiding the
>following problems with Crit:

I don't think it's beneficial to focus on Crit or any singular product on
this list.

I think our energies are far better directed towards working on agreeing on
what set of features comprise an "annotation system."

>	- Crit is a bottleneck.  It's wasteful of resources, and therefore
>	  slow for me to fetch local pages using a remote server.

Ping has acknowledged this before. A proxy solution is a good prototype,
but not scaleable to many users, and not compliant with conventional URL
space.

>	- Crit is a single point of failure - if the Crit server goes down 
>	  I can't see /any/ annotations.

The cost/benefits of redundant server solutions is not unique to
annotations-- most Internet services have to address this problem.

>	- Crit doesn't support distributed databases, "ratings" or other 
>	  things that would be necessary if the service became very 
>	  popular.

These are among the known features of the annotation systems.

>Actually I have lots of ideas. :-) If I had time this is what I'd be
>hacking on, but for now all I have time to do is share my thoughts.  My
>basic idea is to use a slightly modified Crit as a back end for storing
>and creating annotations, and creating a local proxy (similaur to the
>Junkbuster or WWWOFFLE) to merge the annotations with the web page
>itself.

Mind your jargon.

Your "back end" is commonly known as an "annotation server."

Your "local proxy" is commonly known as... well, we still don't have a
name. I'd suggest "client helper", it's the place where the annotations are
collected and prepared for the user; "local proxy" sounds too much like
"server proxy" (one instance of that being the CritLink Mediator).

I'll get to working on that glossary. If anyone wants to help, but doesn't
have time to code, the least you can do is help out in defining the
terminology. Not only do we need good names and definitions, but we need to
recognize all prior names.

Jon
Jon Garfunkel ............................... 
Software Engineer ................................. 
GTE Internetworking /Powered By BBN/ ......
Burlington, Mass ...........

Received on Friday, 27 August 1999 13:09:53 UTC