I assume this implies that we’re not going to mix the marketing release
name (e.g., “Helios”) with the engineering version number, and keep the version
numbers unconstrained. It may be the best compromise, but I have a concern
that’s similar to the one John states below. A lot of people still consider
the version of the Galileo coordinated release to be 3.5. It was tweeted as #eclipse35.
When I get asked about Mylyn version by those who aren’t Eclipse insiders I’ll
find myself saying “Mylyn version that shipped with Eclipse 3.5”. The ECF
download page states: “ECF 3.0 Install via P2 Repository -- for use with Eclipse 3.5”.
I always make an effort to use “Galileo” and proper version
numbers, but then end up saying things like:
“Make sure that you’re on Eclipse
3.5, with Mylyn 3.2, WTP 3.1 and ECF 3.0.”
Which tends to result in a confused stare as someone tries to
pick their way through our 3.x version soup.
For users that just want the latest thing, we’re in a pretty
confusing state due to the various version number, so the naturally fall back
to the version of the Platform. The fact that so many of the numbers are 3.x
just adds to the confusion. (That seems be a point of convergence for API
stabilization.)
Taking the user’s point of view, this has made me secretly wish
that we had a coordinated opt-in version number for projects, pegged to the
Eclipse Project’s. Projects that are big enough to effectively market their
version number, such as BIRT and WTP, could opt out. For integrators, bundle/API
version numbers would evolve as they do now. In other words, just as platform
versions its components separately from the coordinated 3.5 version number, projects
would still use their own internal versioning, but could advertise their
release version as part of Eclipse 4, then Eclispe 5, and avoid the scenario
where a few years down the road we’re trying to communicate a dozen different
3.x and 4.x versions.
Mik
From: eclipse.org-architecture-council-bounces@xxxxxxxxxxx
[mailto:eclipse.org-architecture-council-bounces@xxxxxxxxxxx] On Behalf Of Chris
Aniszczyk
Sent: July-02-09 11:37 AM
To: eclipse.org-architecture-council
Subject: Re: [eclipse.org-architecture-council] The Art of Project
Release Naming
On Thu, Jul 2, 2009 at 11:04 AM, John Arthorne <John_Arthorne@xxxxxxxxxx> wrote:
I can across an
anecdote recently that illustrated the value of release version numbers. I was
shopping for a cordless phone to replace my old 5.8 Ghz phone. When I
last bought a phone, the choice was between DECT 2.4 GHz and DECT 5.8 GHz, with
the larger number generally being a better technology. This time around, there
was a new option: DECT 6.0. After doing some research, I discovered that this
new phone should actually be DECT 1.9 GHz, since that is the spectrum it
operates on. However, the marketing guys thought that a smaller number would
indicate an inferior product to customers, so they pulled the "6.0"
number out of thin air. As a customer I confess I did immediately assume the
higher number was the better technology. Although the conventions are slightly
different, I think this is a perfect example of why a release "marketing
number" can and sometimes should differ from an "engineering
number". It also illustrates the value of a number over simple names (if
they were DECT "zebra" and DECT "camel", a customer
standing in the store wouldn't know which one to pick). All this to say that I
think it would be valuable for us to describe some recommendations on release
naming for projects. Some consistency across projects would be great.
I think the Planning Council will name Eclipse releases monotonically now. What I mean by this is that this year it was Galileo, next year it will be Helios and the following year it will be something that starts with an "I"
There's order implied there so it's easy to talk about things and understand when they came out.
--
Chris Aniszczyk | EclipseSource Austin | +1 860 839 2465
http://twitter.com/eclipsesource
| http://twitter.com/caniszczyk