W3C home > Mailing lists > Public > www-ws-arch@w3.org > January 2003

RE: Myth of loose coupling

From: Assaf Arkin <arkin@intalio.com>
Date: Wed, 8 Jan 2003 17:15:11 -0800
To: "David Jacobs" <djacobs@mitre.org>
Cc: "Ugo Corda" <UCorda@seebeyond.com>, <www-ws-arch@w3.org>
Message-ID: <IGEJLEPAJBPHKACOOKHNMECKDAAA.arkin@intalio.com>

Quantify "little" ;-)

If you look at a service like Amazon you will find a back-end system that is
able to process purchase orders that are transmitted in machine readable
form. That back-end system is utilized by a front-end system that makes the
order process most appealing to humans by breaking a single 50-field
purchase order across multiple pages. A presentation layer is then used to
style these pages into HTML documents.

For machine-machine interaction the most efficient path is to add a
Web-service that talks directly to the back-end. You can deconstruct the
purchase order into multiple HTTP operations on the client side, so they can
be reconstructed back into a single purchase order on the back-end side. It
appears to me that being able to directly send the purchase order to the
back-end through an HTTP gateway would be easier to accomplish for both the
client and the server.

arkin

> I agree that things can be simplified when designing for machine
> reading.  But wouldn't it be nice to get your site machine accessible
> for very little work and then you can work on optimizing it to your
> hearts desire.  Of course the amount you do that will depend on how much
> of your client population is machine based, how much the lack of
> optimization causes additional machine load and additional work for the
> clients.
>
> David
>
> Assaf Arkin wrote:
>
> >>I would think existing web application builders would find this
> >>much easier
> >>to wrap themselves around.  And in fact make it much easier to
> support the
> >>web site for humans and machine agents at the same time.  This
> also solves
> >>the problem or workflow coordination for web services because the web
> >>services tells you when you can execute different methods (just
> >>like it does
> >>for people).
> >>
> >>
> >
> >If you designed an Amazon that was intended for machine consumption you
> >would make different design choices. A lot of the information there is
> >useful for human readers but not quite useful for machine. Some of the
> >information is presented in a particular way to make the page more
> >appealing, improve perceived response time, etc.
> >
> >For example, if a book has ten reviews Amazone would only list two on the
> >first page. But if a machine was reading the information you
> could cut down
> >all the fluff (images, tables, redundant links) and provide all
> ten reviews
> >in one page that is actually smaller than the one containing two
> reviews +
> >fluff. The software would then be able to retrieve all ten reviews in one
> >HTTP request.
> >
> >You would also simplify the steps required to make an order. A lot of the
> >steps were introduced to assist people, but if you automate the
> process than
> >a lot of the text/options are no longer necessary and you can
> cut it down to
> >a single page.
> >
> >arkin
> >
> >
> >
> >>Am I missing something fundamental?
> >>
> >>David
> >>
> >>
> >>
> >
> >
> >
>
>
>
>
Received on Wednesday, 8 January 2003 20:16:15 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:12 GMT