W3C home > Mailing lists > Public > w3c-wai-ig@w3.org > January to March 2003

Re: Amazon Access

From: Al Gilman <asgilman@iamdigex.net>
Date: Mon, 24 Feb 2003 08:03:28 -0500
Message-Id: <5.1.0.14.2.20030224075439.02434b10@pop.iamdigex.net>
To: phil potter <p.potter@chester.ac.uk>
Cc: "Isofarro" <w3evangelism@faqportal.uklinux.net>, "wai-ig list" <w3c-wai-ig@w3.org>

At 07:35 AM 2003-02-24, phil potter wrote:
>As a first time contributor I would have thought that the expectation to 
>get the same results from the same search term is a very reasonable one, 
>regardless of the type of site in use.
>
>Suppose an unsighted user, using the accessible site, needed to pass on 
>search terms to a sighted colleague, using the standard site, so that they 
>could find the same set of results...
>
>This strikes me as being another good reason why it's better to run one 
>site that works for all.
>
>Phil "first timer" Potter

Yes, it is absolutely reasonable for a first time visitor to think this.

No, it is not reasonable for the accessibility community to demand this.

So, once again for the gang:

The user gets the same 181 books in either case.

The difference comes in how these 181 hits are segmented into partial reports.

The decisions that affect how many and which of these hits show up on the
first screen are reasonable to change when the user does it himself with a
machine reader or with a human reader who can scan a longer list of hits
quickly.

The two intermediaries have different performance characteristics and it is
appropriate for the service feeding the intermediary to act differently in a
way that improves end-to-end performance independenly in each case 
(differing delivery contexts) without decreasing accessibility.

Accessibility in this case is the fact that you find the 181 hits and can
navigate to any of them.

Yes we care about usability and performance.  But it is better to give the
best performance under the changing circumstances by changing the behavior;
rather than the same behavior that is best in one case and not best in the
other.

Al

>On Monday, February 24, 2003, at 11:09  am, Al Gilman wrote:
>
>>
>>At 02:04 AM 2003-02-24, Isofarro wrote:
>>>I would expect to get the same search results no matter which front-end I
>>>used (since Amazon is largely a database driven website), but this seems not
>>>to be the case.
>>
>>This is an unreasonable expectation.  To be able to get to all the offerings
>>is a basic access requirement.  That the search should identify the same
>>full bag of hits is probably a corrolary of this.  That different previews
>>with different shortList sizes should always show the same top three; this
>>is not an accessibility requirement nor is it really a reasonable demand.
>>
>>The filtering to get the subset shown first is based on guessing, and it is
>>not unreasonable for the guessing to be sensitive to the size of the subset
>>that fits in the particular view and the likely user preferences that go
>>with variant user demographics per view.
>>
>>To demand the first three in a first page that shows only three to be
>>identically the same as the featured three on a first page that shows ten is
>>neither necessary, nor on closer examination is it in the user interest.
>>
>>Al
>>
>>PS: one controversial source of differences: sniffing
>>
>>People can debate the ethics of the "mixed initiative" priorities that are
>>applied to what comes up at the head of the list, here.  Even on Google you
>>get the sponsored links first.
>>
>>The merchant's objective is to present you with something that you will
>>recognize as something you want to buy.  Before you tire of the chase.  Real
>>web wanderers are fickle, remember the conventional wisdom about how "if it's
>>not visible after three clicks, they're not going to find it."
>>
>>So the site will apply everything it knows about you to guess your
>>preferences as to which of the hits you are more likely to want to buy.
>>These are just guesses, if you put clear logical guidance in the search, the
>>system will honor this.  But you can't totally fault them for including all
>>available information when they are left to guess.
>>
>>The hypothesis that it's more in the user's interest to have the site 
>>guess than
>>not to let it guess is something that I've never seen a scientific study on.
>>But I wouldn't wager against it.
>>
>>If you went to Amazon, you didn't want something that you wanted to buy 
>>to have
>>lain buried in the back room never passing through your "three clicks" of 
>>visit.
>>So consider a bit before you expect total lockstep between views with 
>>different
>>information capacity and user-cost-per-entry-scanned characteristics.
>>
>>Al
>>
>>>From: "David Poehlman" <poehlman1@comcast.net>
>>>
>>> > The address for the web site, discussed below is:
>>> > http://www.amazon.com/access
>>> > and it is easy to use, and nice and clean.
>>>
>>>This is a good example of the disadvantage of having a "text only" copy of a
>>>website for accessibility purposes. Apart from the argument that text-only
>>>versions enforce segregation there is the underlying problem of keeping two
>>>sites up-to-date.
>>>
>>>For example, from http://www.amazon.com/ and enter "Accessibility" as the
>>>keyword.
>>>Do the same for http://www.amazon.com/access
>>>
>>>For Books, the former returns:
>>>
>>>1.) Maximum Accessibility: Making Your Website More Usable for Everyone
>>>2.) CARM: California Accessibility Reference Manual
>>>3.) A Basic Guide to Fair Housing Accessibility: Everything Architects and
>>>Builders Need to Know About Fair Housing Act Accessibility Guidelines
>>>
>>>while the latter returns
>>>
>>>1.) Maximum Accessibility: Making Your Website More Usable for Everyone
>>>2.) Web Accessibility for People With Disabilities
>>>3.) Pocket Guide to the ADA: Americans with Disabilities Act Accessibility
>>>Guidelines for Buildings and Facilities, Revised Edition
>>>
>>>
>>>I would expect to get the same search results no matter which front-end I
>>>used (since Amazon is largely a database driven website), but this seems not
>>>to be the case.
>>>
>>>Recalling from Joe Clark's book (which I'm reading at the moment) "Separate
>>>pages are not equal".
>>>
>>>
>>>Mike.
Received on Monday, 24 February 2003 08:03:51 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 19 July 2011 18:14:08 GMT