Archive of UserLand's first discussion group, started October 5, 1998.

XML-RPC over HTTP or Something Else?

Author:fisherm@tce.com
Posted:10/26/1999; 12:12:48 PM
Topic:Response to Flutterby
Msg #:12371 (In response to 12344)
Prev/Next:12370 / 12372

Let me try to provide some perspective here, as a member of the IETF HTTP Working Group. Another group wanted to print over HTTP, a protocol now known as Internet Printing Protocol (IPP). From what I gathered from reading the HTTP mailing list, the IPP people were only the largest and most vocal of several groups wanting to run protocols on top of HTTP. This led to the Internet-Draft Dave featured in Scripting News about how to decide whether a protocol should be layered on top of HTTP or exist as a separate protocol. (A look at ftp://ftp.isi.edu/in-notes/rfc2568.txt, "Rationale for the Structure of the Model and Protocol for the Internet Printing Protocol" might be instructive here.) So that Internet-Draft was never intended as a slam to Dave's work, as its creation basically pre-dated XML-RPC.

That said, I think there are good points to be made in that draft. However, I've done some work on designing a network protocol (Event Notification Protocol, a lightweight protocol similar to interrupt-driven I/O in the hardware world), so I can understand why XML-RPC started out life as a layer on top of HTTP, because I was going to prototype ENP in a similar fashion. It's just that there's a lot of baggage in HTTP (caching and multiple MIME types for data) that I didn't need in ENP, so I figured that eventually it would become its own protocol, although it would look a lot like HTTP. From what I understand of XML-RPC, the situation is much the same for it. I suspect that inside the corporate firewall, there are a lot of my colleagues in IS who would feel much more comfortable about putting an ENP or an XML-RPC server or client on a PC who would not feel comfortable about putting a web server on that PC (we're working on providing content tools so people don't have to administer their own web server).

Last point -- if XML-RPC really does need store-and-forward capabilities, NNTP and SMTP should be studied for clues as how best to do this for XML-RPC. It may be that both types of protocols -- request-response and store-and-forward -- are needed as underlying protocols for XML-RPC, just as in DCE RPC and ONC (NFS) RPC.

Hopefully I've shed more light than heat here... ====================================================== Mark Leighton Fisher Thomson Consumer Electronics

There are responses to this message:

This page was archived on 6/13/2001; 4:53:12 PM.

© Copyright 1998-2001 UserLand Software, Inc.