W3C home > Mailing lists > Public > ietf-http-wg@w3.org > April to June 2004

Re: Proposal for an HTTP ERR method

From: Henry Story <henry.story@bblfish.net>
Date: Thu, 24 Jun 2004 16:16:40 +0200
Message-Id: <1CC07C88-C5E9-11D8-8703-000A95D9FA7A@bblfish.net>
To: ietf-http-wg@w3.org, Atom Syntax <atom-syntax@imc.org>

In order to avoid an avalanche of emails generated by responding point 
by point to each of the thoughtful answers, I would like to see if I 
can summarize the main points.


1. There is consensus that automated error notification systems are 
needed.

They are needed because the specs now require broken content to be 
discarded but market pressures will requires them to nevertheless be 
interpreted. In a war between standards and market pressures, market 
pressures win. Automated error reporting _in_a_form_to_be_determined_ 
will create a feedback loop allowing a wider number of actors to help 
cooperate in the application of the standards.

2. The question of who/what should be notified of an error.

This is a secondary question.

Here I have a number of contradictory intuitions.
    - If an object in the day to day world breaks, say a tire, I don't 
notify it that it is broken, but the manufacturer. Perhaps the tire has 
a "Service-Error: xxx"  phone number printed on it that I can use to 
send a problem report to the manufacturer
    - If I am talking to someone and they are stepping on my toe, I 
might say "err, you are stepping on my toe". I may nevertheless 
continue to listen to what they are saying.
    - If I send a government agency a request for a form, and they send 
me something back in spanish (which I don't speak, but can decipher due 
to the similarity between Spanish and French, which I do know), I may 
send back another letter to the same address explaining the error in a 
language well understood by everybody to be an error message. And if 
this is not the case in your country, would you not be pleased if it 
were the case? Perhaps one could all agree that letters sent in a RED 
envelope are error messages to the agency to which it is addressed. It 
would then be a matter of the institution of which the agency is a part 
to define a policy of what to do with red letters. They may allow them 
to go to the agency itself. They may sample some of them to see if the 
agency is adapting its behavior correctly. They may send them to some 
completely different address that specializes in dealing with those 
messages.

For me as a consumer the red letter method is much easier to work with. 
I know that the first letter reached the agency (whether the address is 
a PO box that gets forwarded to someone else or whatever is none of my 
concern). So out of all the things I know from that institution to be 
alive, that one is the one I have the most direct evidence for being so 
- I am holding the evidence in my hand! Notice that I do not care that 
my letter reaches the exact same person (server, apache instance, 
thread, processor) who dealt with my previous request. I am speaking to 
the agency resource, and as far as I am concerned that is all I need to 
know.

The relation between these intuitions and the two proposals are quite 
clear.
  - PacerErrVerb: http://www.intertwingly.net/wiki/pie/PaceErrVerb
  - PaceServiceError: 
http://www.intertwingly.net/wiki/pie/PaceServiceError


Henry Story


PS. For the Europeans among you who know about Asterix and Obelix, 
there is a cartoon called "Les 12 Traveaux d'Asterix", where the two 
heroes are given the task to get an answer for a simple question from a 
building. Simple enough they think. The problem is that the building 
turns out to be inhabited by a sprawling bureaucracy where each office 
they turn to, an unpleasant person just redirects their call to some 
other floor and other office. 
Received on Thursday, 24 June 2004 10:16:47 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Friday, 27 April 2012 06:49:31 GMT