Re: [PATCH 2] count-objects: add human-readable size option

From: Petr Baudis <pasky@suse.cz>
Date: Aug 14, 2008, 16:34 UTC
Message-ID: <20080814163446.GD10360@machine.or.cz>
In-Reply-To: <48A45CC9.5040708@griep.us>

On Thu, Aug 14, 2008 at 12:26:49PM -0400, Marcus Griep wrote:

Show 6 quoted lines

> Petr Baudis wrote:
> > Are you aware of progress.c:throughput_string()? It would make sense to
> > use the same code in both instances.
> 
> I was not. After reviewing it, it is limited to its purposes, but
> consolidating the human-readable-ness is a good idea.

Of course, it would need an usage overhaul. But otherwise, it seems fine? Terabyte-sized objects in Git would be very troublesome venture for many reasons.

> (who's got transfer speeds in TiB/s?)

Maybe Dana Brown? ;-)

Show 14 quoted lines

> > I'd prefer you to keep using binary units instead of the ambiguous
> > prefixes, since we should keep our output consistent and I believe they
> > usually end up to be the least confusing choice. (Otherwise, don't you
> > want to use "bkM" instead of "BKM"? I never really know.)
> 
> In general, "b" would be supplied as a part of the suffix, so that is no
> longer in the prefix list.  The distinction comes with kilo vs. kibi.  In
> an earlier email reply, I mentioned a flag to denote SI versus binary
> periods.  In common nomenclature, Kilo (1000) is designated 'k', while
> Kibi (1024) is designated 'K' (the 'i' after the 'K' is supplied by the
> suffix if desired).  Thus, if the user wants binary, they'll get the capital
> 'K', and if they want SI, they'll get the lowercase 'k'.
> 
> Sound reasonable?

I'm confused - you didn't seem to really address my suggestion. Is there a good reason _not_ to go with the /.iB/ prefixes, and just forget about SI? Who is ever going to need SI?

-- 
				Petr "Pasky" Baudis
The next generation of interesting software will be done
on the Macintosh, not the IBM PC.  -- Bill Gates