Hi, Jim
On the one point, I don't personally think using _Orion_ means a lot of
trial-and-error.
We just got to beta with a J2EE app that consumed about 50 man-months, not
counting requirements analysis. All development was on Orion, and
pre-production testing is on Orion. Production will almost certainly also be
on Orion, so Evermind is going to get a few dollars soon. :-)
I'm team lead on this project so I have decent insight on what exactly
consumed a lot of time. And what that was was J2EE itself. There aren't many
J2EE experts out there, myself included, and when we nailed down high-level
design half a year ago there were even less. I think we avoided a lot of
gross errors, but at the detailed design and implementation levels there was
a lot of "well, let's see if this works" kind of stuff.
Orion's friendliness to development (never having to use GUIs is a big one;
allowing exploded deployment structures is another) allowed us to bring
junior developers up to speed very quickly. We've now got a team who, to a
man, can write XML deployment descriptors from scratch, and therefore
understand deployment and configuration - they're much better
troubleshooters, and now I'd be willing to let them use GUI deployment tools
(because many big app servers have nothing but). The same
development-friendly features gave developers super-fast turnaround times
during unit and integration testing: we use Ant and Jikes, and with close to
200 source files (and about 400 other files, mostly XML and XSL) we can
re-compile, redeploy, and immediately see change effects in under 30
seconds.
I estimate that if we had had to use another server in development and
testing, that one could multiply total time required by at least a factor of
1.5. That's a healthy chunk of cash. :-)
To sum up, I believe strongly that the major time consumer on J2EE apps is
J2EE itself. Any realistic app finds itself eventually using BMP entity
beans with data access objects (since, across-the-board, we're not at CMP
2.0 yet), trying to figure out how to implement honest-to-God application
services within J2EE, wrestling with session management (HTTP + session
bean), experimenting with and then extricating from the unholy mess that is
exemplified by the Petstore front-end, and on and on.
And, you know, we've only actually managed to break Orion in one way -
giving it bad XML config files (it's own, not application-specific ones).
Everywhere else it's been very robust. Sort of compares nicely with using a
GUI development tool from another vendor, and 5 minutes into entering values
into attractive JWidgets, having the tool break. :-)
On another note, our network admin (who's never seen J2EE before) had IIS
working as a front-end for Orion, and clustering and load-balancing
happening, all within 2 days. Just working from Orion docs and with our
input. Pretty damned good if you ask me - we haven't even figured out how to
deploy a J2EE app with JRun yet. :-)
I guess I'm sounding a bit like Kevin Duffey. :-) Orion just happens to be a
serious productivity tool when it comes to J2EE development, IMO.
Arved Sandstrom
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]On Behalf Of Jim Archer
Sent: Friday, October 20, 2000 10:05 PM
To: Orion-Interest
Subject: RE: Orion in production - Let's sell support!
Hello all...
--On Friday, October 20, 2000 5:35 PM -0300 "Juan Lorandi (Chile)"
<[EMAIL PROTECTED]> wrote:
> And also, lack of support & documentation is becoming now, as most
> developers are finishing
> their work and reach deployment time(from what I pick up of many mails in
> this list), a critical point about orion.
> Many of us are reaching the point where we have to prove no only that
> orion's the best, but that it also is a good business
> choice. This is unfairly hard due to little colaboration from Evermind's
> team regarding, as said, support & documentation,
> tough it clearly seems to be changing.
This is a key issue. There is an old saying that time is money. Not true in
software. Time is far more valuable than money. Money can be raised but
time can not. Orion is reasonably priced for the product itself. However,
if using Orion means a lot of trial and error development and no official
support from the vendor, the costs in extra consumption of developers time
and oppertunity loss from delayed market entry could easily exceed the
price tag of Weblogic.
Don't get me wrong, I like Orion. I like it alot. Currently, our intention
is to complete development on it and then license it and deploy with it and
hopefully sell it with our product. This goal would be one heck of a lot
easier to obtain if we had official support from the vendor. Right now,
there are people here banging their heads on the wall just trying to guess
at what works and what dossen't, whats implemented and whats not. It's
tireing.
Anybody want to help me start a business selling Orion support on a 900
number? Just charge several dollars a minute, on an incident by incident
basis. If the support is competant, it would sell big. Heck, we could make
more money then Evermind! <Big Grin> OK, just kidding, but this is a
serious issue.
Jim
