[tex-implementors] Re: [tex-live] LM as the default outline font?

Philip TAYLOR P.Taylor at Rhul.Ac.Uk
Wed Mar 30 11:15:10 CEST 2005

This whole discussiom has become considerably more
heated than seems appropriate for this list, so
I would like to start by apologising to anyone
who feels that I have been over-confrontational :
I kicked in in defense of Staszek's original
thetorical statememt and question ("Don't know
why 0.982 is [an] "unfortunate version number".
Should all the world follow Debian conventions?")
and since then things have rather gor out of hand.

However, it does seems to me that the discussion has
raised one very real issue : in version numbers
with a single period separator, is the second
element intended to be a real or an integer ?
Clearly most believe the latter, but I remain
unconvinced that this is universally accepted.

If we look at the evolution of TeX82 prior to
V3, we see that Knuth adopted the following
version numbers :

	0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9,
	0.91, 0.92, 0.93, 0.94, 0.95, 0.96, 0.97, 0.98, 0.99,
	0.999, 0.9999, 0.99999, 0.999999,
	1.0 (I stop at this point)

Now, /if/ Knuth believed that the second element was an
integer, why did he not use V0.10 after V0.9, then V0.11
and so on ?

I argue that his numbering unequivocally indicates that
he was treating version numbers as reals : the use
of an increasing number of significant digits,
as in 0.99, 0.999, ...0.999999 is clearly intended
to indicate the closeness of the release to the
(future) announcement of a putative V1.0 (which
eventually we see).  In order to indicate this
closeness, Knuth (and anyone else who wants to
use the version number to convey more than
simple rank--ordering) is /forced/ to treat the
version number as a real, since (of reals and
integers) only reals are capable of indicating
asymptoticity : 0.999999 is clearly "very close"
to 1.0 only if it is perceived as a real; if it is
perceived as an integer, then it is no closer to
1.0 than is (say) 0.1000001, 0,123456789,
or any other increasingly long string of digits.
Indeed, the very concept of being "close to"
1.0 is meaningless if the "1" and the "0" of "1.0"
are treated as orthogonal dimensions rather
than a single-dimensional real.

At this point I stop and rest my case : I hope
that those who felt that I was either playing
devil's advocate or playing politics will now
accept that I was actually trying to raise a
genuine point of semanhtic interpretation, and
that we can restore TeX-Implementors to its
previous state of calm and equanimity.

** Phil.

More information about the tex-live mailing list