No subject

[XCS TAP]

             Comments on the LIA (draft of August 31 1992)

                             W. Kahan
             Computer Science Division and Mathematics Dept.
                     University of California
                           Berkeley, CA

The LIA is a cynical attempt to pass off a private bill as if it were a public
service when in fact it will do the public a severe disservice.  The latest 
draft offers a proforma but not material response to the objections raised
over earlier drafts, as if the draft's authors felt no embarrassment when 
their tendentious rationalizations contradict their own statments of aims in 
the draft's first pages.  The draft is couched in a dense notation as if it 
were some kind of Mathematics; but in fact it is a pseudo-mathematical 
pastiche, imposing restrictions here and relaxing them there without any 
obvious reason apparent through the fog of symbols.  There is an unobvious
reason for the authors' whimsy; the promulgation of the LIA as an inter-
national standard would redound to the commercial advantage of just one 
computer manufacturer.  By cloaking their proposed standard in mathematical 
obscurantism, the authors may succeed in getting it past both that 
manufacturer's competitors and an unalerted public.

The community of producers and users of numerical software, the community the 
LIA purports to serve, has not yet awakened to the threat posed by the LIA.  
Such protests as have come from that community have been disregarded by ANSI 
X3T2 and the ISO for lack of technical expertise in an admittedly mysterious 
technology.  Thus those bodies are relegated to the status of a Court of Appeal
preoccupied solely with procedural matters since it cannot pass on a lower 
court's illogic nor on the validity of new evidence.  Thuse, ANSI X3T2 and the
ISO are posed to inflict harm they cannot understand upon a community they 
intend to help.

A better procedure would be to send the draft back for a considered re-
evaluation by its prospective victims.  At the very least, the LIA should be 
subject to unhurried public debate in appropriate forums like ...
	In the USA,	ACM SIGNUM
			IEEE Computer Society
			SIAM

	In the UK,	IMA

	    etc.
Their silence in the past cannot be construed as consent.

Nobody has presented evidence that the LIA's promulgation would enhance the 
quality or availability of numerical software; such evidence as exists points 
to the contrary.  Why rush to force upon us all what only its proponents want?