W3C home > Mailing lists > Public > w3c-wai-gl@w3.org > July to September 1999

Re: Content negotiation example needed.

From: Alan J. Flavell <flavell@a5.ph.gla.ac.uk>
Date: Sun, 19 Sep 1999 15:07:59 +0100 (BST)
To: Nir Dagan <nir@nirdagan.com>
cc: WAI Guidelines List <w3c-wai-gl@w3.org>
Message-ID: <Pine.OSF.4.10.9909191433460.6875-100000@a5.ph.gla.ac.uk>
On Sun, 19 Sep 1999, Nir Dagan wrote:

> I agree. One should use both content negotiation and have links to 
> alternative versions.

It seems we all agree with this, although sometimes the reasons are
different [grin]

> In addition, returning 406 messages is problematic since IE5
> repaces the message with its own "friendly message" that says that
> the server could not preform the request, without listing
> alternatives, leaving the user in a dead end.

This is certainly rude of MSIE, but it's documented that it only does
that when the page data is under a certain size.  So that offers the
first workaround: increase the size of the status-406 response page
(yes, this could call for a modification in the server, with servers
that have the response page coded in the software).
  
A second workaround would be to ensure that one document version has a
higher qs factor than the other versions, which makes it improbable
that the server will (as a result of negotiation) need to offer the
reader a choice of versions.  This can certainly be done with Apache,
although it's more fiddly to set up than the simplicity of the
MultiViews option.

But let me reiterate that I favour also offering explicit links to the
other versions, as already noted.

Finally (although I'm sure I'll be told that users must be considered
incompetent to configure their browsers according to their needs) this
bit of MS sillyness can be turned off in the preferences:

  Tools / Internet Options / Advanced / 
  [ ]Show fiendly HTTP error messages.

It's a great pity that MS don't seem inclined to conform to even the
mandatory requirements of HTTP, before starting to branch out with
(maybe well-intentioned) extensions that can have perverse knock-on
effects like this.  At least the extensions ought to be opt-in rather
than opt-out.

> is better to return an unacceptable response and let the user
> choose whether he wants to read it, to follow a link to an
> alternative or not read anything.

I agree.  But if the "unacceptable" response crashes their browser (as
I'm told some charset advertisements can do, although this isn't
something I am experiencing myself), then sending that "unacceptable"
response is also going to be counter-productive.

As always on the WWW, we are dealing with compromises here.  It's not
surprising that, coming from different directions, we may be applying
the same underlying principles and yet reach somewhat different
conclusions on what the optimal compromise may be.  Which, it seems to
me, confirms the benefits of staying flexible, and offering more than
one way to reach the content.

all the best
Received on Sunday, 19 September 1999 10:08:12 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:47:00 GMT