On Jun 30, 2006, at 4:52 AM, Michael Bauroth wrote:
Assume that I have a protocol, which contains some embedded key- value pairs. The current implemented behaviour decodes this data, convert it to POJO's and forwards it to the SessionHandler, where the data then is processed further. The disadvantage of this handling is, that I need for each key-value pair at least 1 POJO class (and have much construct - destroy - garbage collect + data streaming stuff).Now my question: What do you think from the Mina design point-of- view about the following solution? I decode the key-value pairs in my decoders, store it in the related session hash map and don't forward any ProtocolDecoderOutput to the handler. Only when some data is disrupted or the whole message is received, the decoder "fires" a "message completed" info over the ProtocolDecoderOutput to the handler, so that the handler can process the data, which were previously stored in the session.
You could certainly do this.Just because your filter gets a 'messageReceived' event, it does not mean that it has to pass the message on. Its perfectly valid to batch messageReceive events in your filter until you have a more 'complete' message, and only then pass it on down the filter chain.
-pete -- [EMAIL PROTECTED] - http://fotap.org/~osi
smime.p7s
Description: S/MIME cryptographic signature
