Re: A brief analysis of dynamically generated web pages and

Hi, Nir

Actually, the issue of specifying the tailoring can be handled in a
number of different ways depending on the application.  For example, if
the user is accessing a personal portal, when he/she signs up for the
portal, there could be an option for blind user customization.  In a
search form, there could be a button requesting output customized for a
blind user.  Cookies are another option.

Scott

> One issue in generating dynamic pages tailored for the user,
> is the ability of the server to have useful information on 
> the user's needs.
> 
> It seems to me that the current state of HTTP doesn't allow 
> to get enough information regarding needs of disabled people.
> 
> The mechanisms available now to the server are:
> 1. The client's "accept" "accept-language" and "accept-charset" request
> headers,
> 2. The client's "user-agent" request header.
> 
> The accept-language and accept-charset are the most used by 
> clients and servers, but they have no direct relation to accesibility.
> 
> The "accept" header may be of some use for accessibility, but if the server
> wants to serve only different HTML pages, its use is quite restricted.
> 
> The user-agent request header has two problems. One is that using it is not
> scalable
> from the server's point of view: the server has to study the particular
> abilities of 
> a growing and diversified set of user agents. The second problem is that
> even if the server
> recognizes the user agent, this is insufficient to tell what the user
> needs: I may very well
> want a page crowded with images when I am using Lynx as I have configured
> an external image viewer.
> 
> Style sheets offer a mechanism to discriminate by "medium". (Either via the
> "media" 
> attribute of the <link> and <style> HTML elements, or internally in the
> stylesheet itself)
> 
> However as Scott suggests, sometimes complicated transformations (not
> available in CSS
>  but possibly available in XSLT) that are taken on the server side are
> desireable.
> 
> It seems that a possible solution is to have a "user-agent-media" request
> HTTP header.
> (or "accept-media" ??) As it has a limited set of values it is much more
> scalable 
> than user-agent, and is a more natural way to apply style sheets on the
> server side. 
> 
> Seems as an idea that fits the current slogan "CSS on the client and XSLT
> on the server."
> 
> Regards,
> Nir.

Received on Friday, 19 November 1999 22:42:04 UTC