Re: [Medup] Issues draft-birk-pep-trustwords (was: New Version Notification for draft-birk-pep-trustwords-03.txt)

[Christian Grothoff <[e-mail]>]

On 3/22/19 11:55 PM, Bernie Hoeneisen wrote:
> Dear MEDUP list
> 
> The IANA Registration for trustwords (a basic concept applied in pEp)
> https://www.ietf.org/internet-drafts/draft-birk-pep-trustwords-03.txt
> has several open issues, we'd need your feedback on, e.g.:
> 
> 
> 1) Localization:
>    How should non-ASCII-7bit letters or language symbols be represented
>    in the wordlists? UTF-8? HTML-like encoding? IDNA? Any better solution?
>

UTF-8. IDNA/HTML add unnecessary horrific complexity. Also, with IDNA
you are stuck on Unicode 6.0 "forever".

> 
> 2) Registration format:
>    The format for registrations of at IANA is normaly (peudo-)XML.
>    While XML is concise for registering wordlists, the files
>    for registration will contain lot of XML overhead and thus get rather
>    long. Is this an issue? Use another format, e.g. CSV?
>

I don't see it as a major issue, but it might be nice to offer
alternative formats as a convenience (or scripts to convert to CSV/SQL).

> 
> 4) Bitsize (how many bits can be mapped with a wordlist):
>    e.g. a bitsize of 16 allows for a list containing 2^16 = 655536 words.
> 
>    Should this be kept open or should only certain values for bitsize
>    be allowed to be registerd?
>    If latter applies, which values are useful (e.g. 8, 12 and 16)
> 

Any size is technically usable, and if I have a usable list with say
2^14 words it makes no sense to require me to restrict to 2^8 or expand
to 2^16.  So I would definitively not limit to any particular number of
bits. However, maybe a minimum makes sense, like at least 2^8, to ensure
that the generated trustword-sentences are somewhat sane (otherwise some
joker might define a 2-word list for a language...).

My 2 cents

Christian