Re: alternative interlingua

Danny Bobrow <bobrow@parc.xerox.com>
Message-id: <IakYJeQB0V0wA1R40O@nero.parc.xerox.com>
Date: Thu,  9 Aug 90 22:26:02 -0700 (PDT)
From: Danny Bobrow <bobrow@parc.xerox.com>
To: interlingua@vaxa.isi.edu,
        pfps@allegra.tempo.nj.att.com (Peter F. Patel-Schneider)
Subject: Re: alternative interlingua
In-reply-to: <9008091817.AA04788@vaxa.isi.edu>
References: <9008091817.AA04788@vaxa.isi.edu>

Peter,
  Your list of proposed features seems ambitious to me, but also seems
to cover the right initial territory.  However, like Ramesh, I don't
understand your objections to extension facilities.  

I think language extensions are useful, and necessary if we are to keep
this Interlingua alive.  I think that we need to specify some kernel
that all systems can understand (or perhaps a set of layers of
understanding).  Systems are welcome to extend the language provided
they specify necessary conditions in the kernel (at some level).  If
these are also sufficient conditions, then this extension is only for
linguistic convenience.  If not, then the residue of the meaning of the
extension (e.g. with respect to indexing, say, or control) can be
described in a paper or a code or whatever.  All inferences that could
be made at the kernel level are preserved, but (possibly new) systems
can implement special code to take advantage of the residue meaning of
the extension (or whatever part of it they desire).   This means that
two Loom systems can communicaate in something close to full Loomese,
and if Classic is close, it ought to be easy to adapt at least some part
of the Loomese. 

The essence of what I am saying is that the communication between two
systems using the interlingua should be able to move as high as the two
interchanging systems have common ground.   It also lets us (modulo
syntax) agree on some piece of a kernel before we have to agree on all
of it.


----
In your note you make the claim that it would be hard to pass
information between Loom and Classic in the current interlingua.  It
would be interesting to see one example of where such difficulty arises,
and how using the proposed set of features this difficulty is
ameliorated.

danny