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

Philip TAYLOR P.Taylor at Rhul.Ac.Uk
Tue Mar 29 18:54:20 CEST 2005

Gavin McCullagh wrote:


> The point is simply one of convention.  It's impossible for software
> authors to guess how many subversions they will need so rather than having
> to pad the first version 0.00000001, they say 0.10 > 0.1 because 10 > 1.

/Some/ software authors may say; others may say that if 0.1 has
been released, then there will never be a V0.10 (although
there may well be a V0.11).  These same authors would
happily accept that that 0.2 > 0.1 /and/ 0.2 > 0.11

> Subsubversions are done with an extra full stop.  This full stop is not a
> floating point, it's just a delimiter.

Yes, I have no problem with that : V0.1.1 > V0.1
However, would you assert that V0.1.0 > V0.1 ?
If so, on what basis ?

However, Staszek is not using such a scheme, it would
appear to me : he is using a simple real number,
expressed to whatever number of digits seems to him
and his co-autho0rs to best reflect the degree of
difference from any preceding version.
> As an example of this, the man page of the unix ls command says:
>        -v     sort by version 
> this implements the above scheme:

Is perhaps even predicated on it.  However, what the Unix "ls"
command does is neither here nor there : it may be of interest
to Unix users, and it may well be beneficial for Unix users
to number their packages in accordance with a scheme with which
"ls" can cope, but that is surely of no concern to people such
as Staszek et al, implementing /portable/ software which need
pay no lip service to any one operating system's conventions ...
>   gavin at ravioli gavin> touch v0.99 v0.982
>   gavin at ravioli gavin> ls -lv v0*
>   -rw-rw-rw-  1 gavin gavin 0 2005-03-29 17:22 v0.99
>   -rw-rw-rw-  1 gavin gavin 0 2005-03-29 17:22 v0.982
> implying 0.99 < 0.982. 

In the Unix world.  Nowhere else, as far as I know, and
certainly not according to the generally accepted
principles of real arithmetic.

  While of course everyone is in a position to break
> conventions, it is not generally a good idea.

But as you yourself point out, this is a Unix convention.
Staszek is not, as far as I can tell, offering anything
that is Unix-specific.

> Knuth's numbering scheme is consistent with the above.  He just skips an
> awful lot of possible version numbers. 

For all $v$, $v \in <the set of possible TeX version numbers>$,
v' > v follows the rules of real arithmetic.  Knuth doesn't /skip/
version numbers, he restricts himself to a well-defined subset.

> Even LaTeX does this with its sectioning commands.

Yes, with embedded period : as I said above, I have no problem with
such a numbering scheme.

** Phil.

More information about the tex-live mailing list