fun with last-minute, major changes.

[Dave Kormann <davek@research.att.com>]

so, after talking with vijay on friday, i've made some signficant
modifications to the spec.  only one semantic thing, but it's yet more
new or modified text:
- i changed 'signed integers' to 'unsigned integers' in section
2.4.1.  are there objections to this?  in particular, should we allow
negative version numbers?
- i eliminated the text

<P>If the server receives no reply to the <SAMP>mcp</SAMP> message, it
may either attempt to use MCP 1.0 or assume that MCP is unavailable
altogether.</P>

from the discussion of version negotiation.  this is a potential
blockage, and now that we have the new version: and to: args for
#$#mcp, we're protected in the 1.0-communicating-with-2.1 case.

- added a block diagram (WOO, PRETTY PICTURES) of vijay's sketch of
the implementation structure.  pleasepleaseplease filfre to comment on
the text around this.  it's pretty bad right now, and i'm VERY open to
changes.

these are more substantial changes than i made before.  please glance
at the spec and let me know what you think.
				dk