Re: Keeping HTML Simple & Format negotiation between Browser & Server (Jonathan Abbey)
Date: Fri, 28 May 93 10:56:52 CDT
From: (Jonathan Abbey)
Message-id: <>
Subject: Re: Keeping HTML Simple & Format negotiation between Browser & Server
---------  Received message begins Here  ---------

In response to:

> From Thu May 27 15:33:24 1993
> Subject: Re: Keeping HTML Simple & Format negotiation between Browser & Server 
> In-Reply-To: Your message of Thu, 27 May 93 14:24:29 CDT.
> Organization: Berkeley Software Design, Inc.
> Date: Thu, 27 May 1993 15:08:29 -0500
> From: Tony Sanders <>
> My prefered scheme for selection of links in external objects is just send
> the x,y coordinates of the selection to the server and let it deal with
> it however is best.  This is how the current selection stuff works for
> <IMG ISMAP>.  This lets you backend with whatever you need (like maybe
> WAIS spatial indexing) without hardcoding anything in the client.
> This scheme has the advantages of being easy on the browser, simple to
> implement, low network overhead, it fits in with the existing HTTP
> protocol (even HTTP/0.9) and it's working today:

It's easy on the browser, but devilishly difficult on the server.  With a
list of bounding rectangles for a link, it is possible for the links to be
saved in a static file along with the document they apply to, without any need
for an active process to mediate the clicks.  This way, hyperlinked images can
be stored on, e.g., ftp servers, or cached locally.  

In addition, with the bounding rectangles associated with a particular link,
the browser doesn't need to contact the server to follow a link, and it becomes
possible for the browser to provide feedback when the pointer is moved onto a link.

A list of bounding rectangles need not be onerous on the network, particularly in
comparison to the size of the images to begin with.

> The same scheme could work for any external format with a little
> cooperation from the viewer.

Ditto the bounding rectangles scheme.

A generic point and click scheme could be significantly better for things like
mathematical graphing systems, where the server can do a reverse map of the click
to a point in some transformed space.  This could be a good thing.

Perhaps we could have both systems?

Jonathan Abbey
Applied Research Laboratories
University of Texas, Austin