Congratulations on your paper re: KnowNow

Weiquan Zhao weiquan.zhao@cs.unisa.edu.au
Tue, 16 Apr 2002 11:44:31 +0930


Hi, All

Thank you vert much for your comments and I'm quite interested in seeing
this discussion going on. Here's some points I sense from it so far.

First to answer Mike's questions. I feel that the emergence of KnowNow's
technology have answered them itself.
1) We are not doing research on what an "appropriate" infrastructure for
internet-based distributed applications should be. The fact is simply "the
web is there and we HAVE to use it". "Web browser-http-web server" is the
most used paradigm and we haven't considered other clients as KnowNow did
more for application integration. We are not intended to do anything to the
underlying technologies(infrastructure), say writing a new browser.(to
futher that point, why do we need the web at all to do all of this.) In
addition, we think the fundamental problem is with HTTP, (we have seen
capability of browsers increased pretty much or we can have a "single-page
application" if don't care performance), which we cannot help, poor
http-ng...

2) The mismatch we think is actually due to the requirements of the
applications, which is reflected in what developers have get used to. Nobody
complained about the web in the very first days. We are aiming at a result
that enables developers design from top-down while technologies support from
bottom-up. An architecture is supposed to be such a bridge. Of course it
relies on good implementation technologies for better results. One principle
here is to let the web do what we want, not we do what the web wants.

Besides KnowNow we also looked into other technologies like
Barracuda(www.barracuda.com) and dual-MVC(IBM T.J.Watson Research Center).
It seems that to overcome the weakness we bring in enhancements but as well
as restrictions, such as use of frames(the noframe compaign...). The other
concern is how we can protect against hostile browsers to ensure application
integrity won't be damaged by message from client-side, either intensionally
or not. I found Java Pet Store can be crashed simply by a single "refresh".

The weakness, at the same time the beauty of web based applications is
deployment on-the-fly. With the emergence of SOAP I think there is going to
be a change. If we need the beauty, then we use "pure" web based
applications, SOAP/Web services can take over otherwise.

Thanks again for your comments and they are most welcomed.

Weiquan

----- Original Message -----
From: "Mr. FoRK" <fork_list@hotmail.com>
To: "David Kearney" <David.Kearney@unisa.edu.au>;
<weiquan.zhao@cs.unisa.edu.au>; "Gioiosa, Gianpaolo - GIOGY001"
<GIOGY001@students.unisa.edu.au>; "Rohit Khare" <Rohit@KnowNow.com>
Cc: "Richard Taylor" <taylor@ics.uci.edu>; <FoRK@XeNT.Com>
Sent: Monday, April 15, 2002 4:43 PM
Subject: Re: Congratulations on your paper re: KnowNow


> I took a closer look at that document and have a few comments:
>
> It seems there are two major assumptions in the paper that I consider to
be
> questionable.
> 1 - "the web consists of web browsers (using HTML and other technologies),
> web servers and the HTTP protocol."
> To me, "The Web" is more than just web browsers. The web browser is a
simple
> way to build many kinds of applications. When it fails to provide what you
> need - like "full control of their execution sequence" - then you write a
> new client that provides that capability.
>
> "The client component, normally in the form of web page, can perform only
a
> small portion of a task, usually related to user interaction."
> In my mind, a client can be any kind of code that uses HTTP, hyperlinking
> and GET/PUT/POST/DELETE. Browsers are optional.
>
> The paper seems to be pointing out problems with existing browsers, not
the
> architecture of the Web itself. So... write a new browser.
>
> 2 "Developers must take this mismatch into consideration"
> "It does not match the layering scheme that suits traditional multi-tiered
> applications, and thus poses a challenge to
> developers who are used to the traditional tiering scheme."
> The other assumption that I have a problem with is the idea of letting
> developers do what they are used to. The Web is different than traditional
> tiering schemes. Get used to it.
>
>