Re: Moving VAX into 21 century :-)

[Thor Lancelot Simon <tls%panix.com@localhost>]

On Tue, Aug 27, 2019 at 08:08:49PM -0400, Paul Koning wrote:
> 
> The cleaner answer would be to do what Python does in a large number
> of other places, which is to make some of the methods be available
> or not depending on the environment.  Currently the math.xyz methods
> that related to INF and NaN are not documented that way, but it seems
> plausible enough that they could be.  And if so, the test suite would
> simply deal with that as it deals with any other platform dependency.

It's hard to see it this way when there seems to be no other running
system in the world I could build Python on that doesn't support IEEE
math with INFs and NaNs.  I suppose it's (just barely) concievable
someone still has a 1990s S/390 out there running an AIX VM with hex
floating point only, but who believes Python would be buildable in that
environment?

I think we are really the (extremely) odd one out on this.  If we want
modern languages, we probably have to live with their assumption that
floating point is an IEEE world.

> So I'd call this a Python defect.  The one remaining question is whether
> the Python community would accept a patch for it.

I'm guessing not - to them, it's going to look like breaking their own API
and introducing substantial complexity for their consumers for what is
effectively no benefit.  Also, for us this just kicks the can down the
road -- if we ever run into any Python software that deals in INFs or
NaNs we'll have to patch all of _it_ instead.

-- 
 Thor Lancelot Simon	                                     tls%panix.com@localhost
  "Whether or not there's hope for change is not the question.  If you
   want to be a free person, you don't stand up for human rights because
   it will work, but because it is right."	--Andrei Sakharov