Mailing-List: contact ant-dev-help@jakarta.apache.org; run by ezmlm
Message-ID: <0bf401c07b4e$01b9c030$020a0a0a@alsatian>
From: "Jason Rosenberg" <jason@squaretrade.com>
To: <ant-dev@jakarta.apache.org>
References: <B680FDED.1D60D%jon@latchkey.com>
 <3.0.6.32.20010110170324.00a189e0@alphalink.com.au>
 <3.0.6.32.20010111010337.00a61d70@alphalink.com.au>
Subject: Re: The RIGHT Direction for ANT
Date: Wed, 10 Jan 2001 16:38:47 -0500
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit


> At 12:39  11/1/01 +1100, Conor MacNeill wrote:
> >>
> >> So far your the only one who believes it is misguided. You are also the
> >> most ignorant and have the least experience with Ant and other build
> >> processes - is this telling you something ? ;) If you want to add
> >> procedural elements feel free to. I doubt you will get CVS space here -
> >> mainly as it is a different project from Ant - but sourceforge.org will
> >> host it. I think you will find that most of these elements will be rarely
> >> useful outside small builds thou.
> >>
> >
> >Hmmm. Pete, I'm not really happy about this response. 
> 
> sorry ;)
> 
> >Whilst I usually
> >don't agree with Jason's point of view, I would defend his right to express
> >it here. Let us make the arguments, discuss the points, etc, but please
> >let's keep ant-dev friendly. 
> 
> It wasn't meant to be unfriendly - quite the opposite in fact ;)
> 
> >Sourceforging people isn't the way to go, IMHO.
> 
> Nope - but I doubts it is going to happen with so many people against it.
> So it is more a practical thing - if he wants then do it - actions speaks
> louder than words - maybe the arguments will have more effect once backed
> up with implementation. I am not saying go away - I know people who would
> love to use the procedural parts he comes up with. However whats the point
> of encouraging what I see as a dead end ? ;) I think he will end up tired
> from arguing with us and thus end up not doing anything. This would be a
> shame.
> 
> Cheers,
> 
> Pete
> 

Please don't worry about my feelings here, all, I'm not overly sensitive,
or anything.

It is important to answer a few of Pete's criticisms, though.  First,
I have a huge amount of experience in working with C/C++ software
projects, and with using Make, etc.  Pete seems to saying that,
since I don't agree with him, I must not have much experience, or
anyting.

It is true, I have come relatively recently to the world of Ant and Java.
It is obvious, a priori, that Make would not be appropriate, for Java.
I think we can all agree on that.

Ant is the only build tool I know of so far that really is intended to
be for Java applications.  Are there any others that I should
be looking into?

Now, I have managed, kicking and screaming, to accomplish what
I need using the basic features of Ant1.2, and my builds are on
the right track.  The reality remains that I have accomplished this
by doing things in a way that is not really the intended Ant model.

It is all now mostly based on javascript, which calls directly into
the ant java classes to execute tasks.  I do things like make direct
calls into the get/setUserProperty method, which is inaccessible
from the Ant XML language.  I have also made heavy usage
of module based property files, which describe the local
classpath and dependency issues for each module, which
then get loaded in via javascript reusable templates.

Dependency checking is done by simple tracking of modules
that have already been visited at run-time, so I don't have 
any of the issues related to trying to pre-compute a DAG
for the entire build where paramaterized templates are being
used.

So, I have successfully leveraged what's good about Ant.  That is
it builds java well, in a platform independent way.  And, it does
provide scripting, via the <script> tag.

So, in summary, it works, I would have preferred not to have
gone the route I did, I think the powers that be on the Ant
project are trying to win the Nobel prize instead of build
something that works, and I look forward to Ant2, which
should address at least some of my concerns.

XSLT sucks, by the way.

Jason