Lists Home |
Date Index |
On Thu, 2002-02-28 at 15:35, Paul Prescod wrote:
> > HTTP is an RPC application built on TCP/IP.
> If you want to say that HTTP is an RPC *application* then that's very
> different than saying that HTTP is an RPC protocol. It's as different as
> RSS's relationship to XML. They are totally different levels.
Looking from the bottom of the stack up, they don't look very
different. HTTP's relatively easy extensibility (you can conduct RPC
easily in HEAD requests with custom headers, for instance) suggests a
serious blurring to me.
Is REST an RPC application? An RPC architecture? I'd suggest it's the
latter, built on the former.
> HTTP certainly could have been defined as an RPC application except that
> the "RPC protocol" that it builds upon is not documented and not used
> for any other application protocol. So it is at best a virtual RPC
It builds an RPC implementation on top on TCP/IP. The specification
defines the procedures and how to call them. Nothing virtual about it.
> > ... The fact that the method
> > calls are defined *in the standard* doesn't change much about that, and
> > HTTP certainly allows for extensible parameters in any case.
> You didn't touch my question about FTP, or SMTP or POP for that matter.
> You send the UIDL "method" and get back values. Is POP RPC?
I'd say so. Looking at it from the bottom of the stack, definitely.
> > I can certainly define standard non-extensible protocols which happen to
> > use XML-RPC or SOAP.
> But those are not RPC protocols. They are applications of RPC. For
> example, UDDI is not an RPC protocol.
I think you're splitting hairs here to preserve a very odd notion of
REST as pure of RPC contamination. Applications of REST are
applications of RPC. Let it go.
> I'll repeat that HTTP could certainly have been defined on top of an RPC
> protocol without having suffered any harm. It wasn't, probably because
> there was no appropriate RPC protocol available at the time. So it isn't
> the case that everything that RPC touches immediately falls apart. It's
> rather the case that the RPC layer is not a very important or helpful
> layer which is why you can get away without defining it formally. The
> RPC layer only takes on mythic proportions when people start to revel in
> the ability to invent their own protocols on top of it...and that's
> where interoperability starts to degrade.
HTTP could have been defined on top of a generic RPC protocol without
harm precisely because it uses the understandings, expectations, and
mechanisms of RPC. If you really want to claim it's not RPC, I guess
Ring around the content, a pocket full of brackets
Errors, errors, all fall down!