Re: [COMMITTED] pkl, testsuite, doc: defunit, support for arbitrary named units

[Eric Blake]

On 2/29/20 8:37 AM, Alfred M. Szmidt wrote:
> I think following the lead of GNU Coreutils would be wise here
> (https://www.gnu.org/software/coreutils/manual/html_node/Block-size.html):

I agree - any time we have different GNU programs with a different set 
of ideas on what a suffix translates to, it breeds confusion.

> defunit kB      = 10 ** 3;
> # `K' is special: the SI prefix is `k' and the ISO/IEC 80000-13 prefix
> # is `Ki', but tradition and POSIX use `k' to mean `KiB'.
> defunit k       = 2 ** 10;
> defunit K       = 2 ** 10;

Hmm - this would work for bits.

> defunit KiB     = 2 ** 10;

But this, with an explicit B, is wrong (it is not 1024 bytes, but 1024 
bits).

Do we want k to be kbits or kbytes?

> defunit MB      = 10 ** 6;

This name implies bytes, so it should be 8 * 10 ** 6.  Do we also want 
Mb for 1,000,000 bits?

> defunit M       = 2 ** 20;

This is ambiguous on whether it means bits or bytes.  Per coreutils, if 
we want it to mean mebibytes, we need 8 * 2 ** 20.  (Yes, I avoid the 
term mebibytes when I can, and usually say megabytes even when I mean 
2**20 rather than 10**6, but in this context, my choice of wording is 
intentional)

--
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org