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

[Robert Sayre <mint@xxxxxxxxxxxxxxx>]

Mark Baker wrote:

> Hi Robert,
>
> On Tue, Jun 22, 2004 at 01:48:27PM -0400, Robert Sayre wrote:
>  
>
> > We could say "ping". All that's necessary for this to be useful is for 
> > the request to be recorded in the access log, along with all the other 
> > GETs. No processing is necessary.
> > No resource is affected,
> >    
>
> The state of the error log is changed.
>  
What error log? Is a GET that alters an access log unsafe? Most web 
servers do this. Is this something we are concerned about preventing? 
These don't need to go anywhere special. You can run AWStats or whatever 
at the end of the month and find them.

> > the feed is not removed, etc. Requests to the 
> > ErrorURI are also idempotent.
> >    
>
> It's safety I'm primarily concerned about here, not idempotence.
>
>  
>
> > > Here is another test for GET: Do you want a web robot to fetch
> > > this with a GET? When it does that, is there a resource it can read?
> > > Would it be useful to save a copy? Cache it?
> > >      
> > Is a 204 status code allowed for GET?
> >    
>
> Sure, but that's not Walter's point.  His point seems to be that you
> don't want a spider's attempts at indexing your site confused with
> attempts at reporting errors.  I agree with him completely.
>  
That's why the X-Atom request header is required.

Robert Sayre