Le Mardi, 30 sep 2003, � 15:15 Europe/Zurich, Berin Loritsch a �crit :

Bertrand Delacretaz wrote:

Le Mardi, 30 sep 2003, � 14:43 Europe/Zurich, Berin Loritsch a �crit :
...Things like ANT should
be installed like the tool that it is...
I see your point, but IMHO the big advantage with including ANT is that the only requirement on the users is to install a JDK.

We are talking about professional people here. Developers. Hopefully, if they are planning on building Cocoon, they would know enough on how to install ANT....

I know, but since 2.1 (and until real blocks come around), Cocoon *is* indeed distributed in source, see http://wiki.cocoondev.org/Wiki.jsp?page=LatestRelease


There are currently no released builds.

So the problem is currently with users, not developers.

...In projects where I'm working we do it slightly differently, we have a parallel CVS module for tools, shared between projects, with directory names including version numbers, and the build.sh script accesses the appropriate version of the tool with relative paths like
../cvs-tools-sandbox/ant/1.5.1
With avoids heavy duplication of tools yet allows each project to use the "right" version of a tool.

Still -1. While this can be a "special" issue, ANT 1.6 should be backwards
compatible with ANT 1.5.1. I know that 1.5.4 is. They have a number of
test cases to verify that....

In theory yes - in practice, it is very hard to get several project teams to agree on a common version quickly, and during the "transition period" until everyone agrees there can be problems.


But again, I understand your point of view, it is the "duplication of tools" vs. "safer distribution" debate.

Given the current need for a source code only distribution, I think the duplication is "less worse" than the risks of wrong ant versions of installations on the user's side.

-Bertrand



Reply via email to