On 25/09/14 15:07, Christian Schneider wrote:
On 25.09.2014 12:44, Sergey Beryozkin wrote:
Hi Christian
On 25/09/14 08:51, Christian Schneider wrote:
Having one servlet per bundle would definately work.
I see two problems though:
- Each servlet has its own base path where all services reside then. Not
necessarily a bad thing but it would limit the freedom of the user to
define where his endpoints live.
- Added configuration overhead and complexity: Each of the bundles has
to be a web application bundle with a web config and servlet definition.
This is what people do today every day with wars, nothing major and
has proven to work
Unfortunately it is not the same as with wars. Outside of OSGi you
typically have one war per application. So the config effort is only
once per application.
I was primarily referring to your first comment above; to be honest I'm
not trying to start an OSGi vs non-OSGi discussion here, but obviously
people have built complex composite applications consisting of multiple
wars before using whatever inter-war communication mechanism they prefer
such as direct vm/etc
A typical OSGi application often consists of > 10 bundles (without
additional dependencies). So instead of one set of configs you have >
10. So this is considerably more effort.
I thought I suggested a way not to have > 10 configs. I'm fine with you
preferring something different to explore, though as I mentioned I'd
personally not want to embed a servlet name in the URI - but it is only
my preference; subsystems or CXF servlets registered as filters or war
bundles with sharing the configs between them would would work OK for me...
Thanks, Sergey
Christian
--
Sergey Beryozkin
Talend Community Coders
http://coders.talend.com/
Blog: http://sberyozkin.blogspot.com
