Yeah I am in agreement with Mike. An architectural overview...and an
example (heck use AHC vs HttpClient) and show how it scales. That will
get ooohs and ahhhs.
Jeff
Mike Heath wrote:
One of the problems I've had when promoting MINA is that most Java
developers don't understand the scalability implications of the
thread-per-connection architecture. If I jump right in with how cool
MINA is with its separation of concerns, futures, filters, event
mechanism, simplifies packet fragmentation problems, etc., the audience
either gets lost or replies with something like, "Using InputStreamsis
just as flexible as filters but doesn't come with all the difficulties
of having to build a state machine. MINA just makes things complicated!"
However, if I start out showing how quickly I get an OOM exception when
using a thread-per-connection architecture and then show how I can
handle thousands of connections in MINA without consuming loads of
memory, the audience is able to better understand the main problem that
MINA solves. Showing how painful it is to use NIO directly is fairly
simple at this point. The important part is making sure the audience
understands the need for the functionality that NIO has to offer.
Once the audience understands the problems MINA solves, I've found
they're usually much more receptive to the coolness that MINA has to offer.
Just my $0.02.
-Mike
이희승 (Trustin Lee) wrote:
Hi,
I was invited as a speaker of JavaOne 2008 and will speak about Apache
MINA there. Please feel free to contact me to give me some idea about
what you want to hear about MINA if you have any plan to attend this
year's JavaOne.
Cheers,
Trustin
