2:15 p.m.
Copied below is a discussion we had today on #jboss-as7 in IRC that I'm
copying here to get input from the HQ folks.
AS 7 currently has 2 HQ-related subsystems, "messaging" which provides
core HQ, and "jms" which brings in JMS spec compliance. The gist of the
discussion is 1) whether this separation is overly unintuitive and 2) if
it's necessary, is the name "messaging" for the core HQ subsystem too
likely to lead users to assume it's where JMS topic/queue creation is done.
The conversation is below. It carried on after the following; please see
http://echelog.matzon.dk/logs/browse/jboss-as7/1302040800 if you are
interested.
pilhuhn: Also I am astonished that Q s are present , but not Topics
[1:52pm] bstansberry: pilhuhn: there are topics in the jms subsystem
[1:53pm] bstansberry: the "messaging" subsystem is very oriented toward
core HornetQ
[1:54pm] bstansberry: Nihility: ^^^ reminds me of our discussion w/ paul
re: the JGroups subsystem, where we decided it was silly to give a
"generic" name to a subsystem that was not exposing a standard API
[1:55pm] bstansberry: same applies to subsystem=messaging
[1:55pm] pilhuhn: so Qs are in both? messaging and jms? feels strange
[1:55pm] Nihility: bstansberry: makes sense
[1:55pm] Nihility: bstansberry: sort of
[1:55pm] Nihility: bstansberry: another example is subsystem=web
[1:56pm] Nihility: bstansberry: which is jboss web
[1:56pm] bstansberry: well, that at least is exposing the Servlet spec
[1:56pm] dmlloyd: shared global bindings are a nasty little problem
[1:56pm] Nihility: i think our goal should just be to have subsystem names
[1:56pm] Nihility: that are intuitive easy to understand
[1:57pm] Nihility: jgroups vs group-communication is basically the same
to me
[1:57pm] dmlloyd: jgroups is unambiguous and concise imo
[1:57pm] bstansberry: yes, that's decided
[1:57pm] Nihility: yes
[1:57pm] dmlloyd: hornetq should be hornetq
[1:57pm] Nihility: the reason for messaging and jms is different
[1:57pm] pilhuhn: I think it needs to be clear that for an enduse, that
he wants to have a jms/queue and not a messaging/queue when doing jms
[1:58pm] dmlloyd: jms should be jms
[1:58pm] Nihility: they realy do have two subsystems
[1:58pm] dmlloyd: if there is such a generic notion
[1:58pm] bstansberry: right, so calling hornetq "messaging" is not intuitive
[1:58pm] Nihility: they have this notion of a lightweight topic
[1:58pm] bstansberry: people will gravitate toward it
[1:59pm] pilhuhn: Yeah leightweight.. it will float in the clouds
[1:59pm] Nihility: and thats what they think is a real messaging api
[1:59pm] pilhuhn: .oO( DId I win as BS-Bingo? )
[1:59pm] Nihility: jms is just something extra they do for the spec
[1:59pm] Nihility: with extra requirements etc
[1:59pm] Nihility: so its like a higher level layer
[1:59pm] Nihility: so they have core messaging
[1:59pm] Nihility: and they have jms
[1:59pm] Nihility: and togheter thats hornetq
[1:59pm] bstansberry: good point
[2:00pm] bstansberry: ok, no JIRA, at least not from me
[2:01pm] bstansberry: instead, a dev list thread and the HQ people can
deal with it
[2:01pm] pilhuhn: The admin that wants to use a MessageDrivenBean does
not really care about that separation
[2:02pm] Nihility: the separation is more because you can run hq without
starting the jms services
[2:02pm] Nihility: and they have separate configuration
[2:02pm] pilhuhn: yep
[2:02pm] Nihility: although you could collapse the two
[2:02pm] Nihility: emuckenhuber and i debated that for awhile
[2:02pm] Nihility: we had a hard time deciding
[2:03pm] pilhuhn: What I want to express (and that targets rather us
with JON and the console) is that for this said admin it needs to be
totally clear that he wants to go to subsys/jms/queue
--
Brian Stansberry
Principal Software Engineer
JBoss by Red Hat