The Core Problem with Parrot Version Numbers

[chromatic]

On Friday 20 February 2009 09:14:44 Allison Randal wrote:

> The N.N.N versioning is a standard through a significant part
> of the open source world. People will look at it and instantly
> understand "Ah, 2.0 is more recent than 1.0, I might consider
> upgrading." Which is all that really matters.

The *meaning* of those numbers is by no means standard.  (I'm still waiting 
for someone to reconcile the Linux kernel, Emacs, OpenBSD, FreeBSD, Perl 5, 
Mozilla Firefox, Ubuntu, and Apache httpd numbering systems.)

No matter how we choose to number within those constraints, we will always 
have numerologists telling us that we should have meant this or we should have 
meant that.  We will always have debates over what constitutes a major or a 
minor feature, and what breaks backwards compatibility and what doesn't, and 
which upgrades users can skip and which upgrades they should test first.

This is not Sparta.  This is madness.

We already have clear and explained policies for duration of support, release 
processes and timing, and our deprecation promises.  Those are tied to the 
calendar.

If people are going to read that information into our version numbers, our 
version numbers should reflect that.

To make the jump from 2.0 to 2.5 in six months work, we have to say, *right 
now*, that one of the releases will contain only "minor improvements", 
whatever that means.  Is anyone here willing to predict what we'll have ready 
with that degree of confidence a year in advance?  Numerically, this scheme 
*does not work*.  It does not fit how we work, and it does not reflect how we 
release software and the promises we make about future versions of that 
software.

If we absolutely *must* have real numbers as version numbers, I hereby donate 
a nickel to the Parrot foundation so that we can buy more real numbers and 
bump up the major version every six months, thereby solving the "What's the 
deprecation policy?" question and the "Should I upgrade?" question and the 
"There aren't enough real numbers between 0 and 5!" problem.

-- c