On Mon, Feb 28, 2011 at 9:58 AM, ant elder <antel...@apache.org> wrote: > On Mon, Feb 28, 2011 at 9:39 AM, Simon Laws <simonsl...@googlemail.com> wrote: >> On Mon, Feb 28, 2011 at 8:07 AM, ant elder <antel...@apache.org> wrote: >>> On Mon, Feb 21, 2011 at 10:52 AM, Simon Laws <simonsl...@googlemail.com> >>> wrote: >>> >>>> >>>> I wonder whether we need the ability to run from maven at all. We need >>>> a way for people to compile samples of course. I've be happy with the >>>> following. >>>> >>>> 1 - use the shell as the primary out of the box mechanism for >>>> loading/running sample contributions where possible >>>> 2 - present all the other options for running/embedding Tuscany under >>>> "running tuscany" which is after all why we created that directory in >>>> the first place, i.e. to separate sample contributions from the large >>>> number of ways we have of running them >>>> 3 - make using the sample contributions in eclipse really easy >>>> (re-instate the eclipse plugin? provide instructions for generating >>>> projects files? Or even ship project files?) >>>> >>> >>> I can understand the motivation for doing that but we did try it a bit >>> before and I got the feeling people didn't really like it. Eg needing >>> to have a prebuilt binary distribution can be a bit painful. >>> >>> ...ant >>> >> >> Are you suggesting that we don't ship a binary distribution now and go >> with just a source distribution? >> >> Simon >> > > No I wasn't suggesting that, just trying to find some way to make some > progress on the samples. > > It might be worth looking at what the binary distribution is actually > for though, perhaps not in this thread so it doesn't bog down the > sample discussion. The binary distribution isn't used by any of the > Maven builds of the samples or running them. > > ...ant >
I was initially assuming that we would retain a way for users to (re-)compile sample contributions from the source provided in the binary distribution. The complicating factor in these conversations seems to be Maven as it leads to a situation where you download the binary distribution, compile the samples with maven and it downloads all the jars again. Maybe we shouldn't look to Maven for sample support. Maybe we should only allow people to recompile samples from the source distribution but this feels inconvenient as samples without source seems to be missing the point. We could go the other way and do away with the jars we ship with the binary distro and rely more heavily on Maven. I don't really support this approach. Shipping a self contained binary distribution at least gives us some confidence that it will work in six months time. I don't have this level of confidence when using Maven. I recognize that there has been discussion of shipping the jars formatted as a local repo to address this. But you ask a good question. What is the binary distribution for? Three thoughts that immediately come to mind: 1 - A quick start for new users for running samples without the need to compile anything 2 - A vehicle for shipping all the dependencies require by users or embedders of the Tuscany runtime without the need to use Maven to extract them. 3 - An environment were we can reasonably expect other run/build-time environments such as OSGi and Ant to function correctly. I'm thinking that If we could make our distribution/sample/osgi story more accurately/consistently reflect the base + extension story I think the binary proposition, and hence the samples, would be clearer. Regards Simon -- Apache Tuscany committer: tuscany.apache.org Co-author of a book about Tuscany and SCA: tuscanyinaction.com
