Path: utzoo!attcan!uunet!husc6!ukma!tut.cis.ohio-state.edu!ucbvax!HPLB.HPL.HP.COM!rhc
From: rhc@HPLB.HPL.HP.COM (Robert Cole)
Newsgroups: comp.protocols.iso
Subject: Re: Status of ISO RPC
Message-ID: <8907190807.AA00771@rcole.hpl.hp.com>
Date: 19 Jul 89 08:07:42 GMT
References: <8907190644.AA17227@jerry.inria.fr>
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 32

> As ROS appears to many as having pretty much the same purpose as RPC, 
> could you comment on the differences + the rationales for yet another standar
d?

I believe the ECMA RPC standard is effectively a subset of ROS as far
as operational semantics is concerned; though a future standard may
change that.

The two things that an RPC standard ought to add to ROS are:
1. Some idea of a programming language binding, including semantics
that make the use of remote operations obvious and automatic for a
programmer. This includes the notion of binding at compile, link, run,
time which ROS barely touches on.

2. Standard syntax for data structures commonly found in programming
languages and constructs for building the usual complex data
structures. These are all mappable onto ASN.1, but a standard will
give a single way to do this. Arrays and pointers seem to present the
most problems here.

Since the end-point of the protocol is different from that of ROS, and
is more particular, I believe that a separate standard is needed. I do
believe that it will make use of ROS to achieve some of its semantics.
A future standard will have to consider the difference between idem
potent and non-idem potent operations as a possible efficiency
measure. It may also consider the appropriate use of connectionless
working.

These are, of course, my personal opinions and I would be happy to
debate them.

Robert