Thinking aloud : pass-by-ref JMS

[strayalien]

All,

Just some thinking out-loud noise for the forum.

My team is creating a transactional engine based on async messaging. We have our own internal async engine but are speculating about using JMS.

Async operations are in-jvm (all pass-by-reference - speedy), we have tried to minimise interaction with other machines in a cluster.

The JBoss JMS is pass-by-value. We are considering creating a bare-bones cut-down JMS engine (no bells/whistles) to be an in-memory pass-by-reference aync engine. We really can't do with all the serialisation to/from destinations for performance reasons.

Q1: Can JBossMQ be pass-by-reference? (non JMS spec-conformant, I know) and avoid nasty serialisation.
Q2: Can JBoss MDB (etc) use another JMS impl, or is it bound to JBossMQ?

Many thanks,

Greg.

[genman]

All,

Just some thinking out-loud noise for the forum.

My team is creating a transactional engine based on async messaging. We have our own internal async engine but are speculating about using JMS.

Async operations are in-jvm (all pass-by-reference - speedy), we have tried to minimise interaction with other machines in a cluster.

The JBoss JMS is pass-by-value. We are considering creating a bare-bones cut-down JMS engine (no bells/whistles) to be an in-memory pass-by-reference aync engine. We really can't do with all the serialisation to/from destinations for performance reasons.

Q1: Can JBossMQ be pass-by-reference? (non JMS spec-conformant, I know) and avoid nasty serialisation.
Q2: Can JBoss MDB (etc) use another JMS impl, or is it bound to JBossMQ?

Many thanks,

Greg.

[strayalien]

Thanks genman,

That's brilliant. I was having a sniff around that class. I notice you don't recommend cloning the properties. Is the 'trouble' you refer to with the headers is because message sending changes these values? Can you expand on that?

Many thanks,

Greg.