Re: A WWW-annotation protocol (good ideas and bad ideas)

On Sun, Aug 29, 1999 at 05:29:18PM -0700, ping@lfw.org wrote:
> 
> Sorry, could you explain what you mean by "potato-filtering"?

I defined this term earlier in the message - I meant on-the-fly filtering
and censorship of the annotation text itself, performed by the remote 
mediator or annotation server.  The example I took was a mediator admin
who didn't like the word "potato", so he had it filtered from all
annotations.

The Crit architecture doesn't have this problem (although the current
implementation does), since the "mediator" functionality could easily be
local, and thus under complete user control.

[... about XPointers ...]
> times -- and it still seems to be in flux.  The current draft is
> nearly 90 kilobytes in length, and appears to depend on another
> entity, the XPath, whose specification is over 100 kilobytes long.

Ugh.  That sounds terrible, and does not inspire confidence.  I would tend
to agree that something minimalistic would be more appropriate.


On Mon, Aug 30, 1999 at 10:08:30AM +0200, Laurent Denoue wrote:
> 
> Nobody talks about the problems involved with a proxy (local, remote or like
> Crit "mediator").
> They all have one problem in common : they may not work with Web pages
> containing Java applets which
> need to read/write data from/to their original server.
> I don't know if this can be fixed, but I don't see how. Any comments would be
> appreciated.

If the normal HTTP proxy syntax is used, you will have the exact same level
of functionality as people stuck behind a "traditional" firewall/proxy
implementation.  If applets don't work in such an environment then I'd
venture that they were broken by design already, considering how pervasive
proxies are - especially where people are doing Real Work and not just
surfing...

> Also, please note that Crit doens't work with FRAMES.
>
> A last problem which Crit is that it MODIFIES THE URL. You don't have this
> problem with other proxy-based programs.

This is true - but it's an implementation issue, not an architectural one.
The current Crit implementation is a balanced trade-off between creating a
"flawless" system and one that works with the WWW today.  Tighter
integration with the user's browser would eliminate this problem - but that
doesn't mean we can't take advantage of the rest of Crit's current
functionality.

I think Crit would make a really good annotation server, but we should agree
on a standard for client<->annotation-server communication as soon as
possible so better clients can be written that work with Crit, and will also
work with any new servers people come up with.

> I really think that the client-side architecture is the best one. Of course, it
> may not work with all existing browsers, since
> it requires a different implementation in each one. But look at the advantages :
> 
> - you don't have to update your parser to handle annotations on new formats
> (HTML... , XML, others ?)

Why not?  I don't understand how putting the code in the browser suddenly
makes it irresistant to change... ?

[... about 3V ...]
> But the biggest problem is that it works with a central, CONTROLLED server 

Agreed!

> Someone wrote that Crit doens't have this problem : of course it does. Your
> annotations are stored on the Crit Server,

No, that's not mandatory.  Check again - the mediator allows you to
reference documents anywhere online, not just on Crit's local web site.
Crit nicely provides unguaranteed storage to make using the system easier,
but it isn't required.

> There is already a way to annotate the Web : you just write a web page with
> links to the pages you want to annotate.

Exactly - we agree again, and so does Ping.  Isn't that nice? :-)


Someone here (sorry, I forgot the name) was going to draft an XML
specification for client<->annotation-server communication.  I'm looking
forward to seeing it. :)

-- 
Bjarni R. Einarsson                           PGP: 02764305, B7A3AB89
 bre@netverjar.is           -><-           http://www.mmedia.is/~bre/

           This statement is false.  (courtesy of POEE)

Received on Monday, 30 August 1999 05:01:21 UTC