W3C home > Mailing lists > Public > ietf-http-wg-old@w3.org > January to April 1996

Re: Transparency vs. Performance: survey of opinion

From: <kennykb@cobweb.crd.ge.com>
Date: Mon, 26 Feb 96 17:39:35 -0500
Message-Id: <9602261739.AA04476@cobweb.crd.ge.com>
To: http-wg%cuckoo.hpl.hp.com@hplb.hpl.hp.com
Gosh, the misunderstandings under which we all labor.

Nobody, as far as I can tell, is asserting Dan LaLiberte's ridiculous
extreme that a client, once a document is downloaded, isn't free to do
what it pleases with it, subject to the constraints of intellectual
property law and contract.

I worry, however, about leaving it at that.  Certainly, there *are* cases
in which caching is entirely inappropriate.  A simple `do whatever you
please' model is inadequate in the following scenario:

Alice offers a book ordering service.  Bob sees that Alice has a book
for sale, `HTTP Caching in Ten Easy Lessons.'  He fills out a form, and
places an order for a copy to be shipped to him.  The copy goes out via
air express, and arrives on Bob's desk the next day.

Bob peruses the book, and decides it would make a perfect birthday present
for Carla.  He returns to Alice's service, fills out the form again, and
submits it a second time with identical data. 

Unfortunately, David, the system
administrator that runs Bob's corporate firewall, has set a `prefer
performance over semantic transparency' option.   Unbeknownst to Bob,
his second request is handled by David's caching proxy server, and he gets
the response to his first order all over again.

By the way, David
left the company last week to join the new EDI work force at
Dewey, Cheatham & Howe, Inc.  Nobody else at Bob's company understands
Internet gateways.  Also, Edith, Bob's officemate, turned on the
`performance' features in Bob's browser six months ago, so he'd have
gotten a cached response in any case.

Six weeks later, Bob angrily calls Alice and Associates and asks why they
never filled his second order.  When Alice's sales rep patiently explains
that they never received his second order, he rings off and files a
complaint with the Better Business Bureau about Alice's quality of service.
He's printed out BOTH forms that he filled out, and BOTH replies that
he got, after all; `the computer says' that he ordered the book twice.

                          *     *     *

OK, now how does each of the parties propose to resolve this scenario?
Please address the specifics in concrete terms.  I consider several
non-answers to be unacceptable:

 + We can't just say, `educate Bob.'  We WILL have computer-illiterate
   users on WWW services.  Some of them specifically TARGET computer-
   illiterate users.

 + We can't just say `don't implement this sort of application.'  Well,
   we could, I suppose, at the expense of making the World-Wide Web
   irrelevant to a fair fraction of the potential service providers.
   I'd certainly be forced to investigate other means of deivering
   services if the WWW community goes this route.

 + We can, I suppose, require service providers to implement `cache
   busting' schemes.  (In fact, I'm guilty of this, now that Netscape
   is caching POST transactions, and attempting `conditional POST'.)
   Is this our model of the co-operative Net?



--
73 de ke9tv/2, Kevin KENNY   GE Corporate Research & Development
kennykb@crd.ge.com           P. O. Box 8, Room KWC273
                             Schenectady, New York 12301-0008 USA
Received on Monday, 26 February 1996 10:08:52 EST

This archive was generated by hypermail pre-2.1.9 : Wednesday, 24 September 2003 06:31:46 EDT