Went through your (quite interesting) doctoral dissertation and added
some comments inline :-)
Cheers
Prasad.
On 3/30/07, Jason Dillon <[EMAIL PROTECTED]> wrote:
Awhile back I sent some email [1] about restructuring server/trunk
into smaller groups of modules organized by function/feature.
I had been waiting for svk2 to be usable enough to manage
restructuring in a branch while pulling in new changes to src files,
and the latest updates to the svk2 trunk has working support to --
track-renames when merging. Last night I spent a few hours and
whipped up a POC, using svk to move modules around into groups. I've
been tracking changes made to trunk since then and merging them into
my local svk repository and it appears that the --track-rename
feature is working... yay!
I just wanted to provide a little details on this, how it is working
out so far and start up some discussion about eventually making these
changes to server/trunk. Right off the bat, I want to mention that
these changes should probably be implemented *after* we are done with
the bulk of 2.0 work. I don't want to limit this to 2.1, since with
the --track-rename feature it may be very feasible to implement this
change before we are done with 2.0, but should definitely not be done
until we are sorted on the features and TCK muck.
When we do decide to implement something like this, I think we should
also re-groupId things under org.apache.geronimo.server, and use that
namespace for a fresh start... meaning we should not re-groupId to
o.a.g.server until then.
* * *
Below are _examples_ of how modules _might_ be organized, nothing in
stone, probably not completely accurate. I did leave the actual
names of modules as they were, we can deal with the naming of them
later.
So far what I have done was to create 2 new top-level modules:
* framework
* components
These are just pom modules which serve to group other modules. The
'framework' module contains the core (code and configuration) modules
that make up the backbone of the server. Each of these modules only
has dependencies on other modules in this group, or on modules in
testsupport or buildsupport, both of which are built prior to
building framework (except for a wee bit of magic to get the car-
maven-plugin working, see details on that below).
For example:
framework
framework/geronimo-activation
framework/geronimo-client
framework/geronimo-client-builder
framework/geronimo-clustering
framework/geronimo-common
framework/geronimo-connector
framework/geronimo-connector-builder
framework/geronimo-core
framework/geronimo-deploy-config
framework/geronimo-deploy-jsr88
framework/geronimo-deploy-jsr88-bootstrapper
framework/geronimo-deploy-tool
framework/geronimo-deployment
framework/geronimo-gbean-deployer
framework/geronimo-interceptor
framework/geronimo-j2ee
framework/geronimo-j2ee-builder
framework/geronimo-j2ee-schema
framework/geronimo-jmx-remoting
framework/geronimo-kernel
framework/geronimo-management
framework/geronimo-naming
framework/geronimo-naming-builder
framework/geronimo-security
framework/geronimo-security-builder
framework/geronimo-service-builder
framework/geronimo-system
framework/geronimo-test-ddbean
framework/geronimo-timer
framework/geronimo-transaction
framework/geronimo-transaction-jta11
framework/geronimo-transformer
framework/geronimo-util
framework/geronimo-web-2.5-builder
NOTE: this ^^^ is not a complete list, there are still a bunch of
bits in configs/* which I'm trying to figure out where they should
live. See the bits below about framework and javaee stuff.
The 'components' module contains modules for each of the major non-
framework feature components, which in turn contain the (code and
configuration) modules that implement/configure that feature. For
example:
components
components/activemq
components/axis
components/axis2
components/converter
components/corba
components/cxf
components/derby
components/directory
components/hotdeploy
components/jasper
components/javamail
components/jaxws
components/jetty6
components/jetty6-wadi
components/jpa
components/myfaces
components/openejb
components/tomcat6
components/upgrade
components/wadi
components/webservices
As mentioned, each of the component modules contains the (code and
configuration) modules that implement the feature, so for example for
ActiveMQ, we have:
components/activemq
components/activemq/activemq-broker
components/activemq/activemq-ra
components/activemq/geronimo-activemq
components/activemq/geronimo-activemq-management
components/activemq/geronimo-activemq-ra
Where possible, the <dependencyManagement> configuration for
artifacts used by feature components should be put into the
component's pom.xml. For example, the components/activemq/pom.xml has:
----8<----
<modules>
<module>geronimo-activemq-management</module>
<module>geronimo-activemq</module>
<module>geronimo-activemq-ra</module>
<module>activemq-broker</module>
<module>activemq-ra</module>
</modules>
<dependencyManagement>
<dependencies>
<dependency>
<groupId>org.apache.activemq</groupId>
<artifactId>activemq-core</artifactId>
<version>4.1-SNAPSHOT</version>
</dependency>
<dependency>
<groupId>org.apache.activemq</groupId>
<artifactId>activemq-ra</artifactId>
<version>4.1-SNAPSHOT</version>
</dependency>
<dependency>
<groupId>org.apache.activemq</groupId>
<artifactId>activeio-core</artifactId>
<version>3.0.0-incubator</version>
</dependency>
</dependencies>
</dependencyManagement>
---->8----
This isn't going to be possible for all of our dependencies, though I
think that if we can move to this model it would help improve the
maintainability of version information. While that information might
not be in one place anymore, I think that it would help improve
things as it will move the relevant versions close to the modules
that actually use them and thus make management of those version much
easier (as well as making it clear where those deps are used). The
top-level pom's dependencyManagement section is quite difficult to
manage at the moment IMO. I think for the most part we can do this
for most feature components, and for situations where other modules
need those deps, it would be best to have dependent modules depend on
the components/*/* module instead of on the dependency directly, and
if needed create modules simply to provide the dependencies for this
reason.
How about the repositories for these dependencies. Do you envisage
these being split up too or maintained in a single place ? Or would
that become a moot point with the impl of a single (dedicated) repo ?
I also updated the applications/* tree, to give each application in
there its own module for grouping. For example, we had an
applications/console which grouped the console modules, but most
other modules were left directly under applications/*, so I made new
modules for grouping and moved the (code and configuration) modules
under them. For example:
applications
applications/ca-helper
applications/console
applications/dojo
applications/examples
applications/ldap-demo
applications/remote-deploy
applications/uddi
applications/welcome
Just like with components, the code and configuration modules are in
these application group modules:
applications/ca-helper
applications/ca-helper/ca-helper-jetty
applications/ca-helper/ca-helper-tomcat
applications/ca-helper/geronimo-ca-helper
* * *
As mentioned several times above, the code (jar) and configuration
(car) modules are all grouped together. IMO this is key to keeping
the functional components together. It facilitates things like `cd
components/activemq; mvn install` to build all of the components for
a specific feature, as well as allows dependencyManagement to for
feature artifacts to be consolidated into the component groups pom.
Since jar and car modules now can intermingle, that means that the
bootstrap needs to be updated, so that the car-maven-plugin can be
used as an extension. This is relatively easy, we don't really need
stages for bootstrap anymore, just need a single bootstrap profile,
as in:
----8<----
<profile>
<id>bootstrap</id>
<modules>
<module>framework/geronimo-util</module>
<module>framework/geronimo-kernel</module>
<module>framework/geronimo-common</module>
<module>framework/geronimo-system</module>
<module>framework/geronimo-service-builder</module>
<module>framework/geronimo-deployment</module>
<module>framework/geronimo-deploy-config</module>
<module>framework/geronimo-deploy-jsr88</module>
<module>framework/geronimo-deploy-tool</module>
<module>buildsupport/car-maven-plugin</module>
</modules>
</profile>
---->8----
NOTE: I did also move all of the bits from maven-plugins/* into
buildsupport/*, since there are some use cases for modules that
support our build process that aren't really maven-plugins, and maven-
plugins support the build so it made sense to just put them all in
one place.
Then `mvn -Pbootstrap` to get the car-maven-plugin up and running,
which should still only be needed if there isn't a deployed SNAPSHOT
available, else its business as usual.
* * *
The main issues now are really what are the correct components/*
groups and what should be in framework/* and do we need another top-
level module, say 'javaee' to hold the modules which add the feature
non-component modules that support the JavaEE server assemblies? I
think the answer to that is YES, though I think that most of the
modules under javaee/* are going to be configuration modules, like
for example the client.car, this has dependencies to things like
openejb, which are at the moment under components/openejb/*. Other
configs may also fall under that category, where they depend on
framework/* and components/*/* modules.
Previously in the discussion it was mentioned to call this like
'server', though if we change the groupId to o.a.g.server for the
base, then another .server is just going to be confusing.
That reminds me, for the most part these top-level modules are all
going under a separate groupId, so for example, the modules in
framework/* are all o.a.g.framework (or with o.a.g.server as the
base, then o.a.g.s.framework). This makes things nice and
consistent. The only thing which I think we may want to talk about
more related to groupIds is for bits under components/*. One thought
I had was to give each of these their own groupId, though not under
o.a.g.s.components, but as peers to that. For example, the AcitveMQ
modules might use o.a.g.s.activemq as the groupId. This would help
make it easier to match up bits from the src tree to the m2
repository, and reduce the number of artifacts per groupId as well.
I know that some of you might be thinking about that evil windows
path length problem... and its always in the back of my mind...
mostly cursing it for being so dumb, but still its there. And if
that ends up becoming an issue, then I think we should really
consider dropping the org.apache bits from the groupId. But thats
just an idea,
If we ever took this route, then could we put the components under a
"components" name in the groupId ? Eg:
geronimo.server.components.activemq
Just a thought. This will make the framework/* and the components/*
modules consistent w.r.t their groupIds. The groupIds can be mapped to
their source dir layout.
This will also prevent a proliferation of sub-dirs under the
geronimo.server directory in the m2 local repo.
Eg:
geronimo/server/framework
geronimo/server/javaee
geronimo/server/buildsupport
geronimo/server/testsupport
geronimo/server/components
geronimo/server/activemq
geronimo/server/openejb
...
...
...
It would be nice to keep the artifacts of our logical piece called
components grouped together and by themselves. Just wondering out
aloud.
I know everyone else and their mother is using reverse
TLD for groupIds, but many don't have a project in the scale of
Geronimo. I also think the M2 folks didn't even consider this
windows limitation when they moved in that direction else they might
have chosen to implement it differently. Anyways, its minor... but
something to think about. I'd rather we did something different with
groupId's then limit how we can group our modules based on the lack
of intelligent filename handling by everyones favorite love to hate
operation system (okay, maybe its just my favorite :-P).
* * *
So, with all that said, a few hours last night I did play around with
moving bits, and have been sync'ing up changes from server/trunk just
fine into the reorganized tree. I've almost go the geronimo-
framework assembly buildable/runnable. The other assemblies take a
little bit more work to handle updating poms to use new parents to
get configuration correctly as well as to reference modules in the
new groupIds.
Anyhoo... this is just a POC at the moment, though I hope that we can
eventually make this a reality, as I believe it will help simplify
our build configuration as well as facilitate better buildability for
working on specific features.
I'll probably spend a few more hours on this to get at least one of
the javaee assemblies up to finish the POC. The main issue is
figuring out what depends on what and grouping those modules
accordingly. I will probably also merge my local reorg-branch into
something in the sandbox once its functional so that others can have
a look at how its setup. I could have done this already, but I
didn't want to flood the scm list with all of my POC changes, but I
may commit a lump'd merge a little later.
I know its a lot to swallow, but I'd appreciate any comments or
suggestions.
Cheers,
--jason
[1] http://www.nabble.com/Restructuring-trunk%2C-then-next-steps-
tf2175344s134.html#a6014657
