On Sun, 2002-02-24 at 10:51, John Morrison wrote:
> I *know* that it would be the best way forwards - but think about
> the inertia that Sam had to over come to get GUMP accepted (and
> I don't think all have done yet) in Jakarta/Apache projects.

This is the whole problem I see with Gump in that it is currently highly
centralized and not easy for people to participate because it requires
the maintenance of a descriptor that really has nothing to do with
day-to-day development. If you make a project descriptor part of
day-to-day development then participating in Gump builds is no brainer.
This is the approach I've taken with Maven, if you use Maven to for your
project then participation in Gump builds will be automated and correct.

If your project doesn't want to use Maven (or something like it) then
you get to maintain your Gump descriptor by hand and change in
dependencies usually results in a project failing to build until Sam
fixes it manually.

I don't actually care if anyone uses Maven, I made it to help manage
Turbine projects and Tambora, the project I deal with on a daily basis.
But I plan to move all the Turbine projects over to using Maven and
hopefully this will make managing turbine easier and provide up-to-date
correctness for Gump.
 
> Also - how do you see the mechanism for *adding* projects to the
> GUMP build?  

If a project uses Maven then I will use BCEL and a JavaCC-based Java
parser to find all the dependencies and provide CLI and GUI tools for
adding to the Maven descriptor. I intend for it to be automated for the
most part.

> Should we only add the projects which Apache projects
> rely on or should it be open to everyone who asks?  

I don't think so. I'm planning on making it easy to use Maven, and if
someone uses it then they can very easily participate in Gump builds.

> If it's
> everyone have you considered the time it would take to cvs update,
> rsync and build?  Not to mention the harddrive requirements...

Yes, I've considered it. I don't believe building from HEAD on a daily
basis is particularly use for a project. I believe that as a project
gains greater control over its domain and its developers have a great
comprehension overall that they will have time to participate in things
like Gump. For me Maven is not first and foremost a tool of pragmatic
value and I'm not immediately concerned with the social experiment
aspect which is Sam's imperative. I feel cooperation between projects
will happen when the project itself is easier to manage.

So I think building against HEAD is something that could happen on a
weekly basis. Projects could be notified of an impending build, the
descriptor could be transfered to the build machine and then the build
happens. This would obviously need some work but I call it a reactor and
a reactor can have any number of participants. I'm setting on up for
Tambora so we can have a reactor that deals with the projects that are
of immediate concern to me but the same process, I believe, would work
for larger builds. But I feel that the impetus must come from the
projects and it has to be easy to participate which is what I'm trying
to do with Maven.

> I don't see this happening.  Prove me wrong - I'd love to see
> GUMP used everywhere (I use it mainly internally).

I see it happening mainly because I see it as requiring very minimal
effort eventually. I hope to use the Turbine projects as an example of
how easy it will be to participate and at the same time alleviate Sam of
the burden of having to deal with our turbine descriptors which are
often a great source of irritation.

> J.
> 
> <snip/>
> 
> 
> 
> --
> To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
-- 
jvz.

Jason van Zyl
[EMAIL PROTECTED]

http://tambora.zenplex.org


--
To unsubscribe, e-mail:   <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>

Reply via email to