In English, the alphabetical order of numbers is 8, :B, 5, 4, 9, 1, 7, 6, :A, 3, 2, 0. In Spanish, it's 0, 5, 4, :A, 2, 9, 8, :B, 6, 7, 3, 1. This disorder makes it difficult to write software to compare and sort numbers written in word form. This is generally not an issue because, afaik, no software would ever store the number 23 as the string "twenty-three", but would instead use the more economical numeral string "23" or the binary encoding 00010111.

But, in case there's anyone in the world who believes that it would be useful for numbers to be in alphabetical order, here's a proposal for the single-digit numbers:

0 = a
1 = an
2 = bi
3 = drei
4 = far
5 = five
6 = hec
7 = hep
8 = icht
9 = nine
:A = ten
:B = ven

Question: How can this system be extended for numbers greater than *10 or with dozenal points?
Bonus question: What about negative numbers?

Brilliant! Humans will need to try harder given current Billware sorting techniques - we need to handle ranks too. If we are using Bill's software, we will have the additional problem of "fifty-hec" being sorted before "five". We'll also see:

icht
ichthundred-ninety-an
ichtmillion-dreihundred-ichty-dreithousand-ninehundred-tenty-an
ichtynine

The names sort well so long as their ranks are the same, but the rank names will need to be similarly designed in order for alpha sort to handle them. One would perhaps need to state null ranks like an-hundred-aty-a. Also, maybe name lengths need to be standardized. There's a sexagesimal example of a fairly strict alpha regimentation of number words (some of you saw this already) at http://autonomyseries.com/Canon/Sexagesimal/ Scroll down to "Figure 2". That scheme could be redesigned to be more alphabetic.

We have this sort issue when certain folks store files with dates in their names, and use a form like this:

16nov2006presentation.ppt
17feb2006presentation.ppt
17jan2006-100pm.ppt
17jan2006-1100am.ppt
17jan2006presentation.ppt

which is solved if the format is adjusted to the so-called "bigendian" sort (see Wiki).:
20060117.ppt
200601171100.ppt
200601171300.ppt
20060217.ppt
20061116.ppt

This problem is a lot like my project files on the Wintel system here. All my project numbers are dozenal, and I get this:

"r77a2
r77b9
r782b
r7633
r7769
r7795
r7799
r7813
r7817
r7830
r7849"

Wherever I get a "dek" or "el" I have a sort issue, because I use a and b to represent them. Of course this problem would disappear if there were two additional numerals, but then ASCII would need to be rejiggered. ("r" means rendering, next four characters are the dozenal number of days elapsed from 1 July 1970 to the project due date, yes, a measure of time I invented soon after I "discovered" dozenals).

ps, then we will need to teach people how to correctly say "icht", perhaps a lot like what my daughter says when she sees a spider ;-).

Nice as far as it goes, Dan, but to poor a little cold water on this, the current alphabetical order is itself illogical.

The sounds of English, which do not have a direct correspondence to the letters of the alphabet can be put into phonetic order and often are in phonetic charts.

Now, once that is done, we have the additional problem that by virtue of being close in sound, sounds next to each other are difficult to distinguish. For ease of listening, a conlang designer might legitimately choose maximally different sounds. A sort of phonetic equivelent to the qwerty keyboard.