From: Trans <transfire@...>
Date: 2005-10-11T13:21:54+09:00
Subject: Re: facets gem error


Brian Buckley wrote:
> > Okay, please try it again.
>
> Cool.  The error is gone.  One thing however -- the version number is
> "2005.10.11" so for example "gem list facets --local" returns "facets
> (2005.10.11, 0.7.2,0.7.1,0.7.0)" Is the version numbering change
> intentional?

Yes. It is.

In fact, if you don't mind I think it a good opportunity to bring up
some points ont he matter --as I have been debating with myself at
great lengths with reagards to the best course.

This version of Facets is basically a "1.0" pre-release. But as such,
and considering the nature of the project, there will never be any
major compatability changes (with one possbile exception, but even that
probably won't have any major end-user effects). So I must conclude, a
dot-version, as it is generally explained, makes little sense.

I know the Gems people probably detest this, but their claim of the one
and only *Rational* Versioning Policy is ironic for its irrationality.
Granted they need some consistancy across the board to have versioning
be useful, but I think they could afford to be a little more flexible
--like supporting date versions. Thankfully Gems can't tell the
difference between a dotted date and any other dotted number.

So why did I choose to make the switch? Mainly I was tired of picking
numbers out of thin air as if they meant something. If you look at the
changelog you can see waht I mean. Trying to follow along with the
commonly held "rational" modus, I would have been forever gathering
weenys -- v0.9.9999... Hek, I'm rarely even sure when to add a new
minor. And as I said, short of bumping to 1.0, there is no criteria for
ever bumping to 2.0 --which is to say it's all make-believe. In fact I
would argue that no project has ever maintained a "rational" policy of
this nature that was indeed rational for any signifficant length of
time. Great counter examples include TeX which increments it's version
toward lim(PI). And Apache which actaully renamed thier 2nd major
version "Apache2" --yep I have etc/apache and etc/apache2 on my system
L-(. So while one can espouse the theories superiority behind this
so-called "Rational Versioning Policy" all you want, at the end of the
day its just as a good friend of mine used to say: "Great idea! Too bad
it ain't worth a sh*t."

That said, date-versioing isn't perfect either, but a least it clearly
means something. And I've managed a *practical* policy that helps even
more by giving some indication of stability.

  1) Development releases have the full date: YYYY-MM-DD
  2) Stable/Offical releases have just the YYYY-MM
  3) Ultra-stable realeases have just YYYY.

So that's the game plan.

T.