Re: PaceServiceError GET vs. POST (was Re: AtomPubIssuesList)

[Antone Roundy <antone@xxxxxxxxxxxxxx>]

On Wednesday, June 23, 2004, at 01:14  AM, Greg Stein wrote:
> * The ErrorURI and X-Atom approach makes no sense.
I'd have to agree somewhat that GET + X-Atom to distinguish error 
reports from web crawlers and other curious entities feels like a hack. 
 If POST is used instead, then X-Atom isn't needed, right?  The server 
can simply ignore GETs to the URI.  You could do this in Apache 
configuration, for example:

	<Files /filesystem/path/to/error/reporting/uri>
		<Limit POST>
			Order allow,deny
			Allow from all
		</Limit>
		<LimitExcept POST>
			Order deny,allow
			Deny from all
		</LimitExcept>
	</Directory>

> The assumption here is
>   that you want to use existing tools which can GET or POST to report 
> the
>   error. But when you use those tools, they won't necessarily give you 
> a
>   way to insert an X-Atom header. You certainly can't just have a web
>   page do a POST, or hit a link to do a GET. You have to write some 
> custom
>   code to insert the X-Atom header.
>
>   Thus, if you've already axed the possibility of a bare browser, and 
> fell
>   into custom code, then why not use an HTTP method that has actual
>   semantic value? Such as ERR or ERROR.
Tools that don't have the ability to do error reporting for whatever 
reason can be forgiven for not reporting the error.  Plenty of others 
will be reporting it on their behalf.