monotone-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Monotone-devel] Re: Proposal for human readable revision IDs


From: Nathaniel Smith
Subject: Re: [Monotone-devel] Re: Proposal for human readable revision IDs
Date: Wed, 7 Sep 2005 21:18:33 -0700
User-agent: Mutt/1.5.9i

On Wed, Sep 07, 2005 at 09:23:56PM +0100, Bruce Stephens wrote:
> I'm not sure about 27 characters, but I just hacked something to
> produce 14 character strings from an alphabet of 60 characters (I
> think that's about right, isn't it?), and yes, that's much better.  An
> example is "tw7535wMJBCH0U" (the alphabet is upper and lower case
> letters and digits, with O and l removed).

Err, 14 draws from a 60-letter alphabet give you only 14 * log2(60) =
82.7 bits.  You need 160 / log2(60) = 27.09 letters to represent a
full SHA1... (with 61 letters you only need 26.98; since we can't
have partial letters, this comes out to "28" and "27", respectively).

> On the other hand, why do I want the hash to be readable?
> 
> I don't want to be able to remember a hash to type it somewhere else,
> because I'll just cut and paste.  

I do feel the convenience of mildly memorable hashes -- often I'd
rather just retype a few letters instead of reach for the mouse, find
the cursor, position it, curse as my hand slips, etc.  Hashes are
worse at this than many things, because we have no practice at reading
them (as compared to, say, words), I always get "5" and "f" confused
because they sound similar when repeating the digits to myself in my
head, and so on.

More pronounceable hashes do have problems, though -- you don't want
to code them as English words, because 1) that gives you huge
expansion in size, 2) it's totally parochial.  (And 3), in a few years
we'd have to change it to Chinese to match the majority userbase, and
I don't think I could deal with hashes in Chinese ;-).)

You can use gibberish -- bibblebabble's 5-letter pronounceable
strings, I think someone else suggesting
consonant-consonant-vowel-consonant in this thread -- but you run into
the problem Graydon pointed out back the first time we had this big
thread, which is that software the mumbles insane gibberish is not
really that reassuring; and you also run into problems with simply
hitting bad words -- (both bibblebabble and the CCVC suggestion
include a number of scatological, obscene, and otherwise off-color
English words, and this is surely true for other languages as well.
The first time a manager notices that their shipping product's "About"
box says "Build: <male genitalia>"...

(Hopefully managed not to hit anyone's email filter there... :-))

So, dunno, there don't seem to be any really good solutions.

-- Nathaniel

-- 
"...these, like all words, have single, decontextualized meanings: everyone
knows what each of these words means, everyone knows what constitutes an
instance of each of their referents.  Language is fixed.  Meaning is
certain.  Santa Claus comes down the chimney at midnight on December 24."
  -- The Language War, Robin Lakoff




reply via email to

[Prev in Thread] Current Thread [Next in Thread]