W3C home > Mailing lists > Public > w3c-wai-ig@w3.org > April to June 2004

Re: Accessibility of HTML frames in "cloaking" application

From: David Woolley <david@djwhome.demon.co.uk>
Date: Fri, 21 May 2004 08:30:20 +0100 (BST)
Message-Id: <200405210730.i4L7UKF04235@djwhome.demon.co.uk>
To: w3c-wai-ig@w3.org

> To explain:  Various DNS (Domain Name Service) providers and domain registrars 
> offer a "Web Forward" service allowing the use of free Web hosting, such as 
> Yahoo's Geocities, to have a "real" URI.

The ones I have seen use invalid HTML and break some automated tools that
would see through a valid frameset.  The invalid HTML results from having
a frameset with only one frame, but with multiple widths. (It's about three
years since I ran automated tools on one, so hoster may have a better
solution now, or tools may be more forgiving.  I doubt the former, as 
such mechanisms aren't provided for sophisticated users.)

The other major problem with them is that they introduce the fundamental
problem of frames (they are incompatible with a world wide web because
you cannot (easily) bookmark the leaf node pages, but without giving
any benefit, other than the vanity effect.

They are attractive for down market hosting companies in that they only
have to provide bandwidth for a small frameset page accessed once per
visit (less if cached).  Bandwidth is likely to be more expensive than
disk storage.  They are also a special case of an option that uses a
second frame for advertising.

The are attractive for naive but vain content providers because the
the average page viewer only ever sees the frameset URL which does
belong to their vanity domain; however that is also the disadvantage,
as people can only bookmark that URL easily.

Of course, if a search engine manages to get through the broken frameset
page, it will index the real URLs, spoiling the effect almost completely.
I guess some authors might discover boiler plate scripting to force 
deep link entries to the home page, but that makes deep links unusable.

And, of course, the browser history will show the real domain name, not
the vanity one, possibly confusing someone trying to find the page they
looked at the previous day.

I would say that there are basically three acceptable alternatives:

1) the domain hoster forwards the request to the real server and then 
   forwards the reply - this probably doesn't make commercial sense as it
   would be cheaper to simply host the site directly (in particular, it is
   likely to result in images as well as HTML being transferred using their
   bandwidth);

2) the domain hoster does a temporary redirect preserving the local part of
   the URL - typical users will see the real URL in the address bar, but
   can publish deep links with the vanity domain (an author could make all
   their links go through the hosting machine, but that would increase the
   load on that machine and not hide the real name);

3) the domain hoster does a temporary redirect to the home page, preventing
   the use of the vanity form of deep links.

(2) is the best compromise, but may require a technically competent
hoster, and I suspect most cheap hosters are not particularly technically
competent.

Generally, I would say the use of "cloaking" should be discouraged.  For
anyone who has any understanding of the web it looks cheap.
Received on Friday, 21 May 2004 13:15:30 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:13:32 UTC