Re: lynx-dev http://wire.ap.org

[David Woolley]

> i've never before encountered a situation
> in which 2  \'s  take me to a different rendered page.

That could happen if a cache considered the page at least marginally
cacheable and the origin server failed to set a correct Vary header
on a page which was dependent on the User-Agent header.  What might
then happen is that the fastest responding upstream cache varied from
request to request and you got versions of the page tailored for
different browsers - there is an awful lot of ignorance about caching
issues amongst web site designers.

Vary is an HTTP 1.1 header which tells a cache that the content of the
page is dependent on the values of named headers as well as the URL 
requested. A server that is acting on the user agent should set:

Vary: User-Agent

However, most user-agent conditional pages are done using dynamically
created HTML, not content negotiation and, unless, as is unlikely, the,
typically ASP, script sets that header there will be no indication.  Such
pages are largely uncacheable, but an incorrect server date, explicit
expiry information, or an aggressive cache might still result in its being
cached.

An HTTP 1.0 cache would also fail to act on Vary.

I would suggest looking at the headers, and if they don't contain Vary
and/or a combination of Cache-Control: no-cache and Pragma: no-cache,
you should add these points to the complaint about not designing the pages
to fail gracefully when JavaScript is missing of disabled.  Don't expect
any sympathy; they are in business to make money, not to produce valid
HTTP and HTML; you'd need to convince them that supporting Lynx is more
profitable than some other changes they could make.