Hello Robert,
from the beginning of Restlet Framework, our most important aim is to provide a
clear and rich API for developing both client and server side applications.
Implementation comes later and we're fine to take shortcuts and to leverage
well-established frameworks such as Jetty (which was the first HTTP server
connector shipped with Restlet).
Of course, we had, and yet still have, the intention to provide
production-ready home made client and server HTTP connectors. The first
implementations relied on the classic basic IO, then an effort was made to
develop full NIO connectors, because we saw that as an opportunity to provide
scalable connectors, asynchronicity, etc. At a time, we recognized that we were
lacking time or contributions in order to fully achieve this task (even if we
think we are close to it), and thus decided to export this into the
org.restlet.ext.nio module, and replace it with the com.sun implementations
internally. In addition, constantly and transparently we always claimed that
these internal connectors are fine for development environment not for
production.
I'm not sure to give you the best answer, feel free to ask for more, or to help
us reconsider the current position.
Best regards,
Thierry Boileau
PS: I've just committed code in the 2.2 and 2.3 branches, in order to add a
decent thread pool to the internal HTTP server connector.
--
http://restlet.tigris.org/ds/viewMessage.do?dsForumId=4447dsMessageId=3087184
