Re: Support bytesize comparison in sort

[Paul Eggert]

Mart Somermaa <address@hidden> writes:

> As I already explained, the current implementation behaves correctly
> with *both* SI and powers-of-1024 input, so there is no need for two
> option letters.

We need to think of the future, not just for today's implementation.
I expect future users to complain that 'sort' mishandles cases like
1001M versus 1.0G.  At that point we'll need two letters.  (But
perhaps I'm wrong; that happens a lot. :-)

I take your point that my proposed syntax is pretty ugly.  And since I
haven't come up with a better solution, perhaps we should just use
yours.