W3C home > Mailing lists > Public > www-math@w3.org > December 2009

Re: Candidate Recommendation

From: David Carlisle <davidc@nag.co.uk>
Date: Wed, 16 Dec 2009 10:02:00 GMT
Message-Id: <200912161002.nBGA20WI028197@edinburgh.nag.co.uk>
To: andjrob@gmail.com
Cc: www-math@w3.org



> Before MathML3 becomes a true Recommendation, I thought I should point 
> out that the example given in 4.2.1.2 "Strict uses of <cn>" for the 
> "hexdouble" datatype mentions that this holds 64 bits and 16 hexdigits 
> (which is correct), but the example uses 8 hexdigits (which is incorrect).

Thanks for that, will fix.


> On the same subject, I was wondering why the "hexdouble" was created 
A major aim of the revisons in Content MathML was to align with
OpenMath as MathML and OpenMath have always been "nearly aligned"
havig been developed by overlapping groups of people in overlapping time
frames.


http://www.openmath.org/standard/om20-2004-06-30/omstd20html-3.xml#sec_xml

OpenMath (probably) predates C99 design, as  that part dates from some
time in the early 1990's.

Encoding al the forms in the string content without needing separate
markup at the xml level would have advantages and disadvantages. As you
say, it would be a bit simpler for languages that have exactly the same
syntax for double literals, but on the other hand it wuld be harder for
those that don't. If something validates using the (decimal) double form
then you can fairly safely drop it in as a numeric literal in most
languages, however the hex form requires some conversion in most
languages which do not have any native suppor for that. Openmath's
integer markup has this problem (<OMI> 16 </OMI>  and <OMI> 0x10 </OMI>
being equivalent markup for 16) which means that you can not simply
drop an OpenMath XML integer into a conversion unless the target
language supports hex integers, without first inspecting the form of the
string literal. In an xml content it's often far more natural to drive a
conversion from the xml attribute structure rather than the syntax of
strings in the element content.

>  but C99  does have a notation for NaNs, so there is a one-to-one mapping between 
> C99 hex double notation and the binary form.

yes that's the main reason OpenMath has a hex form of doubces as well.

thanks again for your comments and we'll fix the incorrect example.

David

________________________________________________________________________
The Numerical Algorithms Group Ltd is a company registered in England
and Wales with company number 1249803. The registered office is:
Wilkinson House, Jordan Hill Road, Oxford OX2 8DR, United Kingdom.

This e-mail has been scanned for all viruses by Star. The service is
powered by MessageLabs. 
________________________________________________________________________
Received on Wednesday, 16 December 2009 10:04:25 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 20 February 2010 06:13:06 GMT