latest spec

[Erik Ostrom <eostrom@research.att.com>]

I've given Dave my notes on some wording changes and typos and things,
but there was one thing I wanted to bring up on the list.  The spec
defines an algorithm for generating cord IDs, in part:

  2. if this implementation is a server
    1. return concat("I", n)
  3. else
    1. return concat("R", n)

(concat?  why not... string-append?)  Anyway, in this algorithm, "I"
stands for initiator, the initiator of _MCP communications_ (defined
elsewhere to be the server), and "R" stands for the receiver of that
initial #$#mcp message.  I had understood the "I" to be the initiator
of _the session_, i.e., the client, and "R" to be the receiver of that
connection, the server.

I'd sort of prefer the latter model, as (a) our decision about who
"initiates" and who "receives" the initial MCP message still seems
somewhat arbitrary, and (b) "initiator" and "client" seem like the
same thing to me, and it confuses me that here we have them reversed.
(Hello, X Window System.)  But if nobody else cares I'll let it go.

--Erik