On 17-Nov-08, at 8:29 AM, Brett Porter wrote:
I have a number of perspectives on this, and have taken the time
traveling over the last few days to reflect on this decision.
As a mentor, I can recognise the podling in its current form is not
working. Contributions, patches, and questions have gone largely
ignored this year, and development stalled after the migration to a
new trunk. For a long time we have assumed that a kick in that
direction was all that was needed to get it going again, but only
small steps have been taken. Something clearly needs to change.
I don't believe the move you've suggested will make visibility,
community involvement and a steady release schedule any more likely
than it is now - they simply require a conscious decision and group
effort to do so which could equally be achieved here.
All we know is that here did not work. All Shane wants to do is try a
different approach.
My primary role in the project at the moment is as a user,
distributor and occasional developer. There are two reasons your
proposal won't satisfy our needs in that regard:
- we are committed to and required to maintain the code in an
independent environment
- we require integration with Archiva for the search and management
functionality
Do you have some alternative proposition here? I realise these may
not be your personal "itches", but there are no technical reasons
they can't be achieved by other community members that are
interested in doing so while other work continues in the areas
you've described.
I personally agree with the overarching technical direction you've
described. In the long term, I see the .NET portions of this project
doing much of the work independently, and this is also consistent
with discussions I've had with others. However, Byldan hadn't seemed
much more than an experiment to date - could you elaborate on where
you think it stands to be able to provide a solution to Maven users
today, or what kind of effort is required to bring it to that point?
NMaven is useful for people who have hybrid approach: Java on the back-
end and .NET in the front-end. Byldan would be a pure .NET approach
for Microsoft developers who are looking for an analog to Maven
without the requirement of a Java runtime. For a shop that is already
doing Java it's not a big stretch to use Java intermixed with C#. But
for shops doing purely .NET I have found interest in NMaven to be
slight at best. They are almost immediately turned off by having to
install Java anything. Looking at your typical developer using
Codeplex this is the profile that you have. They don't know anything
about Java and don't want to know anything about Java. So NMaven
provides the solution for the hybrid shop, Byldan is for a whole new
and different user base. An attempt to bring a new set of users into a
Maven way of developing software.
I would still urge you to reconsider and give a renewed effort here
to see what we can achieve by openly discussing the issues and
roadmap for everyone that is interested in this project and seeing
what comes of it.
The majority of the IPMC members agree that I think the attempt here
hasn't worked.
However, if you don't believe that's feasible, I would be willing to
champion an effort to try again afresh - either by restarting the
podling (with Incubator and Maven PMC approval), or choosing a new
location, and then migrating the current codebase to something more
suitable with respect to any current users. Hopefully through
regular communication we could converge over time while being free
to experiment as needed.
It can only help to have more people working on it. I wish you luck
with your endeavors.
Thanks,
Brett
On 14/11/2008, at 8:14 AM, Shane Isbell wrote:
I'd like to let the general NMaven community know that we are
dissolving
NMaven from the Apache Incubator and will be moving the project to
Sonatype
Forge (http://sonatype.org) over the next week. I can't speak for all
members of the NMaven PPMC but my personal reasons are that after 2
years in
the incubator, we have failed to gain traction with the community.
The
community involvement was quite active while NMaven was at
sourceforge,
drying up very quickly after moving to ASF.
Looking back, I think there were a number of reasons for this.
First, I
don't think the Apache brand name has as much appeal to the .NET
crowd, so
it was easy for us to get buried. Second, we tied NMaven to a Maven
2.1
snapshot to support Visual Studio, and thus were stuck at a point
where we
couldn't do a release. Finally, I took a divergent path from Maven
implementation, in part to solve problems with lack of toolchain
support and
versionless artifact file names, and this made it difficult for Maven
developers to understand and to make contributions. The 0.15
release earlier
in the year was designed to address these problems. Yet we still
didn't gain
much traction with the user or developer community, pointing to
something
more fundamental.
Given these problems, I'd like to take NMaven to Sonatype Forge,
giving it a
fresh start, getting back visibility, community involvement and a
steady
release schedule. Now in regards to the future plans.
A number of .NET developers I talked to said that they while they
liked
Maven and the concept of the Project Model, they didn't want to
have to
install Java and Maven on their systems to use it. And finding
developers
who know Maven, Java and .NET is not the easiest thing, so I'll be
putting
in some time into Byldan (http://codeplex.com/byldan), a .NET
version of
Maven. I'm looking to port over the new 3.0 project builder code,
so we can
start bringing the Byldan project model inline with Maven's pom. We
can
build the plugins once in .NET and then execute them using either
Byldan or
NMaven. The unreleased version 0.14 has this .NET plugin support,
but that
code needs to be rewritten and/or cleaned up. We will need
developers with
experience in app loading and app domains.
Another area I would like to focus on is the Visual Studio support.
Eugene
has done a lot of work with the m2eclipse plugin, using Lucene
indexing of
the the POMs for artifact searches, so I'm looking to create similar
functionality using Lucene.NET.
The final areas of focus are ones that we have been trying to solve
from day
one: properly handling versionless artifact file names and doing
toolchain
support, at both the client and server level. These problems are
enormously
difficult to do against the web based repositories, so I will be
looking to
integrate NMaven/Byldan with the Nexus Repo Manager Rest APIs, just
pegging
the technology so we can get things moving more quickly. This
should also
make things like PDB attached artifacts easier to handle.
I'll be posting more transition information over the coming week.
Feel free
to shoot back with any questions or concerns.
Thanks,
Shane
--
Brett Porter
[EMAIL PROTECTED]
http://blogs.exist.com/bporter/
Thanks,
Jason
----------------------------------------------------------
Jason van Zyl
Founder, Apache Maven
jason at sonatype dot com
----------------------------------------------------------
What matters is not ideas, but the people who have them. Good people
can fix bad ideas, but good ideas can't save bad people.
-- Paul Graham
