confusing labels

John C Klensin klensin at
Mon Apr 13 02:48:38 CEST 2009

--On Sunday, April 12, 2009 20:09 -0400 Vint Cerf
<vint at> wrote:

> Most States in the US have a similar restriction. I remember
> trying to generate pronounceable passwords, 8 chars long,
> using alternating consonants and vowels. To see how this might
> work out in practice, I generated about 500,000 samples and
> was immediately surprised to find that "DEFECATE" fit this
> formula perfectly. I am sure you can think of others. Imagine
> now trying to find a way to automatically avoid sending to a
> new user his or her automatically generated, pronouceable
> password without accidently sending something really
> offensive. :-)

One needs a dictionary or equivalent offensive word list.  Such
lists are almost possible for a single language written in its
primary script (whatever that means).  But, as soon as one
enters the territory of transliterations, interpretations of
what a string in long script, interpreted with the phonemes of
one language, sounds like to speakers or another language or
looks like when transliterated into another script, one gets
into very nasty combinatorics and likely deep trouble,
especially when one adds in  the factor that one person's (or
culture's) offensive language is another's "free speech" or
legitimate satire.

I'm glad it isn't our problem.


More information about the Idna-update mailing list