Re: Feedback on the ping="" attribute (ISSUE-1)

Mark Baker wrote:
> No.  The implementation is non-idempotent, the message is not.

This depends on some understanding of the semantics of the message, right? 
Which we're in the business of defining...

> Consider that when my Web server receives GET requests, it logs
> (appends) a record of that in a log file.  Does that make the message
> non-idempotent and non-safe?

GET messages are defined to be idempotent in the sense that from the perspective 
of the one making the GET request there should (must?  doesn't matter that much) 
be no difference in the response no matter how many times the request is made 
(modulo the resource actually changing).

 > No, of course not, because GET messages are safe by *definition*.

Agreed, with s/safe/idempotent/, since we haven't even started talking about 
"safe" yet.  And they're defined to be so by the HTTP RFC.

> Until this distinction is appreciated, I don't think we're going to
> find any further common ground.

I appreciate the distinction, sure.  Here are the facts we seem to agree on:

1)  A message can be idempotent or not idempotent independent of how it's
     treated by the server.
2)  Whether a message is idempotent is a matter of definition.

Are we OK on those?

Given that, the specification that defines ping messages is the specification 
under discussion here.  And it defines them to not be idempotent.

Again, this is just the ping message.  This is not the GET generated by the link 
click: that's idempotent.  But the associated ping message, if any, is not.

-Boris

Received on Thursday, 8 November 2007 07:30:13 UTC