Re: svn commit: r773350 - in /tuscany/sandbox/travelsample/chapter-02: client-contribution/ node-jumpstart/ node-jumpstart/src/ node-jumpstart/src/main/ node-jumpstart/src/main/java/ node-jumpstart/src/main/java/scatours/ node-jumpstart/src/main/reso

[Simon Laws <si...@googlemail.com>]

On Wed, May 13, 2009 at 12:31 AM, Simon Nash <na...@apache.org> wrote:
> I investigated this a bit more.  There is a simple change that can
> be made to the launcher code to avoid loading contribution code from
> the classpath.  Instead of the node-jumpstart code using the interface
> com.goodvaluetrips.Trips to call the contribution code, the
> node-jumpstart code can include a compatible interface scatours.Trips
> that is loaded from the classpath.  This avoids the need to load the
> interface com.goodvaluetrips.Trips from the classpath.
>
> I tested this and it works.  I changed the launcher code in the
> jumpstart-launcher directory in the new structure to do this, but
> I did not change the code in chapter-02/node-jumpstart.
>
>  Simon
>
> Simon Nash wrote:
>>
>> I noticed that the launcher in chapter02/node-jumpstart puts the
>> goodvaluetrips-contribution code on the classpath.  In general it
>> isn't a good idea to put contribution code on the classpath because
>> this prevents the classes in the contribution from being loaded by
>> the contribution class loader, which means that import.java and
>> export.java don't work properly.
>>
>> In the launcher in chapter-02/node there is an example of how
>> the code can be factored so that no contribution code appears on
>> the classpath but instead is loaded by its contribution class loader.
>> This requires an additional contribution for the client code, which
>> is separate from the launcher instead of being part of it.
>>
>> Is there a reason why node-jumpstart needs to put the client code
>> and contribution code on the classpath?  If not, I think it would be
>> better to use the pattern in chapter-02/node for all the launchers
>> in the travel sample.
>>
>>  Simon

Ok, that sounds like a reasonable approach. I'll start looking at the
new about the old version.

Simon