Warning headers for offline cache

[David Whitmarsh <da...@dial.pipex.com>]

I've been looking at adding warning headers to pages read from the
cache when offline (see earlier posting re offline cache) as per
RFC2068 section 14.45. However, a quick test shows that IE4.0,
Netscape Navigator 4.06 and Opera 3.51 all ignore these headers. Does
anyone know if anything more recent does? And does anyone have an
opinion as to whether these headers should be sent anyway so as to
conform to the spec, even if the clients don't.


David

************************************

David Whitmarsh

Sparkle Computer Co Ltd

Sybase C C++ perl UNIX NT

************************************

Re: Warning headers for offline cache

[Koen Holtman <Ko...@cern.ch>]

On Fri, 7 May 1999, David Whitmarsh wrote:

> 
> I've been looking at adding warning headers to pages read from the
> cache when offline (see earlier posting re offline cache) as per
> RFC2068 section 14.45. However, a quick test shows that IE4.0,
> Netscape Navigator 4.06 and Opera 3.51 all ignore these headers. Does
> anyone know if anything more recent does? And does anyone have an
> opinion as to whether these headers should be sent anyway so as to
> conform to the spec, even if the clients don't.

Warnings are a bit of a can of worms.  The specification of warnings got
revised since RFC2068, if you do want to implement warnings, take the
specification from draft-ietf-http-v11-spec-rev-06 (see
http://www.w3.org/Protocols/History.html#Rev06). 

The above revised spec will explicitly let you get away with _not_
sending a warning about disconnected operation, though this would be
frowned upon. My opinion is that these headers should be sent anyway by
caches, as a debugging tool if nothing else. 

Some warnings (like 'transformation applied', 'response is stale') must
really be sent if they apply, regardless of whether browsers support them,
because this could make the difference between ordering 1 pizza and
inadvertently ordering 10.  If the cache fails to send such MUST level
warnings, and things break horribly as a result, it will be the fault of
the cache implementers and this could have both legal and puplic relations
implications. 

> David

Koen.