[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160411064802.GB23157@dhcp22.suse.cz>
Date: Mon, 11 Apr 2016 08:48:02 +0200
From: Michal Hocko <mhocko@...nel.org>
To: Alexandru Juncu <alexj@...ux.com>
Cc: "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
den@...nvz.org, redkoi@...tuozzo.com, rkagan@...tuozzo.com,
Johannes Weiner <hannes@...xchg.org>
Subject: Re: [PATCH] proc: meminfo: Replace kB with KiB in output
On Mon 11-04-16 08:35:00, Alexandru Juncu wrote:
> On 11 April 2016 at 08:12, Michal Hocko <mhocko@...nel.org> wrote:
> > On Sun 10-04-16 17:23:11, Alexandru Juncu wrote:
> >> Current values are based on multiples of 1024 (powers of 2)
> >> which means that the values in meminfo are not kilobytes
> >> (1000 bytes) but kibibytes (1024 bytes). The correct
> >> prefix for that would be 'Ki' so the output should be 'KiB'.
> >
> > Does the difference actually matters so much that we should change a
> > user visible format of a file? Some users might not ready to changes
> > (say sombody did sed 's@.*:[[:space:]]*\([0-9]*\) kB@\1@' to get values.
> > This change would break it which is something we try to prevent as much
> > as possible.
> >
> > So I do not think this all is worth the potential troubles. There are
> > probably other places where we present kB while we in fact think kiB.
>
> Isn't it at least worth clearly specifying this in the documentation
> (maybe also comment in code)?
Sure, improving our documentation is highly appreciated of course.
Thanks!
--
Michal Hocko
SUSE Labs
Powered by blists - more mailing lists