commons-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jason van Zyl <>
Subject Re: unmavenising Commons projects
Date Sat, 22 Jun 2002 14:35:35 GMT
On Sat, 2002-06-22 at 01:39, Henri Yandell wrote:
> I'm on a different point of view than costin. I'm +1 using Maven, but a
> mad rush to mavenise all of Commons without structure is going to hurt.

I don't think it's been a mad rush, it's an iterative process. By no
means is it perfect and some builds get bonked when the descriptor is
left behind for a version. No doubt it's a hassle in some cases. But I'm
certain it will make things easier in a matter of months.
> To answer costin first:
> Maven is just another build tool. Why not use it instead of ant in the
> end? Given the great quality of documentation that Maven has, I think it's
> going to be a lot easier for a cvs-user to use than ant currently is.
> It took me ages to get to the point of having ANT_HOME defined and knowing
> to always put junit.jar in my ant-lib directory and to get optional jar
> etc. Maven blows that away. Let's seriously consider a Mavenised Commons.

Again I think the decision has to come from the individual committers
for each component but I do think Maven will make things easier. In
eight weeks time I think it will kick some serious ass. The critical
pieces that are missing are an auto-installer i.e. a user cvsups and if
Maven isn't present it will be installed by asking a few questions, and
second the reactor which is used for processing a massive number of
projects in a defined way. One of those modes will be like Gump. I am
expressly flying down to Atlanta next week to work with Bob McWhirter to
get a full-blown workflow engine started and to work on the Reactor
(which will probably use the workflow engine). Once these two things are
done I hope that it will take all of 30 seconds to get going with Maven
for the first time.

Zenplex, where I work in NYC, supports the use of Maven and has pretty
much given me a hiatus from my normal day-to-day in order to get Maven
off the ground and hopefully by the end of next week with the Reactor
working (and possibly the auto-installer) there will be tangible, very
workable solution.

Also, the project has grown rapidly as I expected because it gives
projects something very valuable without much work. And so people have
contributed back freely because of what Maven provides initially with
little effort. This I think is the critical key. Ensure it does
something for a project in a real way and you will get some buy in. Once
the projects have freely bought in then you can start making some real
progress toward a stable massive build system. IMO, this will never work
otherwise. No inter-project will work without something like Maven, I
haven't seen anything come even close.

> Mavenising a project is stunningly easy once you get into it. A small
> group of committers can mavenise Commons very easily. In fact, one of the
> most annoying parts of Maven at the moment is the build.xml file, which is
> hopefully going away. [I'm sure Jason or someone will correct me on any
> Maven mistakes I make, I'm not an Maven project member]
> Replacing the current Commons website with a Maven generated site is a
> delicious concept. It's so easy. It will be even easier.

Again, I hope that Maven speaks for itself. It is an incredible touch
subject asking be to change their builds but I think it is the job of
the Maven developers to make this a transparent process for users.

> To answer those who are pro going straight to maven right now:
> Maven is not ready. There are going to be big versionning issues. It moves
> at an amazing pace and is changing a lot. I think Commons needs to get in
> on Maven now and look to replace the website, look to replace the build
> and look to give Commons some much needed structure. Commons is a great
> shop window for Maven.
> But Commons should not make itself wholly dependant on Maven. That is far
> too bleeding edge. If Maven slows, or enters a redesign period, Commons
> exists in a limbo waiting for a complete redesign, tied to Maven. Maven
> versions are not backwards compatible, only one version can run per user
> [without some major sneaking], and doesn't have versionned documentation
> yet. And it shouldn't have, it's still at a pre-project sandbox stage,
> despite the high speed and professionalism. We will quite soon reach a
> point where parts of Commons need Maven b3, parts need Maven b4 but kinda
> run in b5, and parts need b5.

I think the generated build.xml that works with vanilla ant to produce
the JAR is a workable solution. But the build.xml file needs to be
standard because building the various commons components is still a PITA
using just Ant in many cases because of the fiddling involved.
> I'm sure it's not acceptable for me to make the Codec project work only
> under Ant 1.5 Beta 2. We're going to be doing the same with Maven.
> I'm sounding like a broken record...
> Hen


Jason van Zyl

In short, man creates for himself a new religion of a rational
and technical order to justify his work and to be justified in it.
  -- Jacques Ellul, The Technological Society

To unsubscribe, e-mail:   <>
For additional commands, e-mail: <>

View raw message