What a character

Here's a "captcha" I failed today:

I failed it when I typed in KeZLNDdM. I presume because the Z was supposed to be a z. Which would lead me to:

Lehmer's First Law of Captcha Design: If you are going to make your captchas case-sensitive, then only use instances of letters where their case is apparent from the character's glyph, not just its size. Which means that you can use D and d, E and e, G and g, but not Z or z.

Which reminds me of an observation I had when we were teaching the kids to read - children don't have to learn 26 letters, but 52, in pairs of characters that "mean" the same thing. Although actually it's less than 52, since some letters look alike (or close enough) in both upper- and lowercase.

A serif font makes the differences even more apparent:

• Aa - 2
• Bb - 2
• Cc - 1
• Dd - 2
• Ee - 2
• Ff - 1
• Gg - 2
• Hh - 2
• Ii - 2
• Jj - 2
• Kk - 1
• Ll - 2
• Mm - 1
• Nn - 2
• Oo - 1
• Pp - 1
• Qq - 2
• Rr - 2
• Ss - 1
• Tt - 1
• Uu - 1
• Vv - 1
• Ww - 1
• Xx - 1
• Yy - 1
• Zz - 1

The numbers above are my own personal reckoning of whether the two versions of a letter are similar enough to be recognizable in both cases by learning only a single character (counted as 1), or are they so different that a child has to remember that both R and r refer to some Platonic R (and hence count for 2). Using that criteria kids have to learn 38 letters to read, but we only teach them to sing 26. It's almost like a cruel joke.