Re: [Nmh-workers] new marker format broke test suite

[Ralph Corderoy]

Hi Paul and Ken,

Paul wrote:
> but if we don't, should the conversion use 1024 ("2.4M") or 1000
> ("2.6M") as the divisor?  and of course that leads to MB vs.  MiB.
> and how many digits should there be after the decimal?  sigh.

If it's base 2 then it must be Ki rather than K, etc.  I don't think the
B is needed and can be assumed?  The user can always add it as plain
text.

Ken wrote:
> I'm fine with 1000 as the divisor (that's what is done now, right?
> No, actually, I'm wrong; the divisor is >> 10).  So I guess 1024 is
> fine as a divisor, actually ... I don't have strong feelings about
> that.  A single letter indicating the metric prefix is fine to me.

Metric implies base 10.  http://en.wikipedia.org/wiki/Binary_prefix has
lots of detail.

Paul wrote:
> as for the name:  i'm not sold on "metric" either.  something like
> "numabbrev", "numunits", or maybe "si_units"?

`si_units' have to be base 10.  `units' suggests the function gives just
an appropriate unit for a size and not the size itself.  How about
`kilo' and `kibi' functions, allowing the user to choose their base.
They obviously cope with more than K/Ki, and would be documented so, but
to me the names are immediately indicative of what they're doing and the
base used.  The shipped files would continue to be base 2 by using kibi.

Cheers, Ralph.