NA and NaN

[Francesco Potorti`]

I see that Octave has NA, which is described in the manual as "a special
case of the representation of `NaN'".  I do not understand what this
exactly means, because

isnan(NA) --> 0

which means that NA is not a NaN.

Unfortunately, the above implies that NA cannot be used with any of the
nan* functions in the octave-forge statistics package.

What is the practical utility of NA?  Is this an old Octave (and R)
usage that has never been exploited?  Would it make sense to change the
nan* functions in the statistics package or those in the nan package to
use NA instead?  Would it make sense to have isnan return true on NA?
Or should one simply forget about NA and use NaN since this is the
common usage?

Shouldn't a discussion on the usage of NA be included in the appropriate
manual page?  If I get a unique answer to the above questions, I
volunteer to write it.

-- 
Francesco Potortì (ricercatore)        Voice: +39 050 315 3058 (op.2111)
ISTI - Area della ricerca CNR          Fax:   +39 050 315 2040
via G. Moruzzi 1, I-56124 Pisa         Email: address@hidden
Web: http://fly.isti.cnr.it/           Key:   fly.isti.cnr.it/public.key