Re: RFC for mediator warnings

> Regarding your proposed extension, I suggest you look at the 
> HTTP extension framekwork that the W3C HTTP team has been working on
> and proposing to IETF recently.

Is there something wrong with the way I'm using then?  It's just that 
I wouldn't call it an extension as defined by that document.  I've now 
read it through and I'm not convinced it will work in this case for the 
following reasons:

1.  Although it's possible to get the gateway to send (for example) an 
unsolicited optional extension in its header, it is unclear how browsers 
may react to this.  Ideally they should ignore them, but I can just 
imagine some browser coming out that reports a message back to the user, 
which is totally not what I want.  Or even worse, fail to distinguish 
between optional and mandatory and treat the page as "500 server error".

OK, such a browser may be non-standard, but it is the nature of my 
access gateway to cater for non-standardness.  I can't take the W3C's 
attitude of "I'm standard and if you're not then tough".  If browsers 
were standard then the access gateway would not be needed in the first 
place.

If a message is intended for gateways then it should be in a form that 
is most definitely not meant for end-user browsers.  The working draft 
is all about end user browsers and I can see this confusing things.  
Neither will hop-to-hop work, since there is no telling who's going to 
be installing a gateway.

2.  I'm a student, not a company.  The working draft specifies that you 
need a permanent URI (or at least, as permanent as the extension).  It 
seems to be geared towards the company itself being the only one that 
uses that extension, or at least if anyone else uses it then they have 
to do so in the lifetime of that company.  I agree that making names 
globally unique is a very difficult problem, but I don't think that URIs 
are the solution, since the web is so non-permanent (especially for 
people like me).  When I wanted a globally unique name, I just picked 
one that it seemed unlikely anyone else would pick (at least, a search 
for X-Mediator-Warning produced nothing).

3.  It's more complicated.  This does not encourage authors of other 
gateways to implement it.

The way I suggested is a bit more like the Standard for Robots Exclusion 
(especially if you do it in META tags).  Do you think I should go the 
whole way and use META NAME instead of HTTP-EQUIV?  (If so, who controls 
that standard?)

Regards

-- Silas S Brown, St John's College Cambridge UK http://epona.ucam.org/~ssb22/

"It is better to debate a question without settling it than to settle a
 question without debating it" - 18th century philosopher Joseph Joubert

Received on Monday, 1 March 1999 02:41:24 UTC