Message-Id: <200207141827.OAA10957@vizcacha.cox.net>
From: radian AT cox DOT net (Adrian Mariano)
Date: Sun, 14 Jul 2002 14:27:14 -0400
In-Reply-To: "Re: ANNOUNCE: DJGPP packages of GNU units 1.80" (Jul 14, 14:08)
X-Mailer: Mail User's Shell (7.2.6 beta(5) 10/07/98)
To: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>,
   JT Williams <jeffw AT darwin DOT sfbr DOT org>
Subject: Re: ANNOUNCE: DJGPP packages of GNU units 1.80
Cc: DJGPP newsgroup <djgpp AT delorie DOT com>
Reply-To: djgpp AT delorie DOT com

On Jul 14, 14:08, Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk> wrote:
> Hello.
> 
> JT Williams wrote:
> > 
> > -: from its readme file:
> > -:
> > -: "GNU `units' converts between different systems of units.  It can
> > -: handle multiplicative scale changes.  It can also handle nonlinear
> > -:                                                          ?????????
> > -: conversions
> > -: such as Celsius to Fahrenheit."
> > 
> > FYI, the readme is broken; Celsius and Fahrenheit are _linearly_ related.
> > Does this merit a bug report?
> 
> Yes, I think so. I've BCC'd the author on this reply.
> 

No, the transformation between Fahrenheit and Celsius is an affine
transformation, not a linear one, because of the constant offset.  If
F is a linear transformation then it must satisfy the property
F(a+b)=F(a)+F(b).  This is clearly false for the above temperature
conversion.  For example 32 degrees Fahrenheit is 0 degrees Celsius
but 32+32=64 degrees Fahrenheit is not 0+0=0 degrees Celsius.

This difference between linear and affine is not a minor issue.
Fahrenheit to Celsius conversion was the most requested "units"
feature for years, but I had no way to implement it because the
fundamental design of the program could only handle linear
transformations.  And furthermore, it wasn't even obvious how to
extend the design in a natural way to make this conversion possible.