Just learned my last software project failed
I just learned that the last job I was on, the project failed. It was one of those large S.A.I.C projects so it was probably expected. I was a lead on the project (maybe that was the reason, hehe).
Here are my thoughts:
Over-designed; business people had way too much control and screen layouts with billions of fields
Too much J2EE technology; we used every project that was out there, I mean literally every part of the J2EE stack (EJBs, JMS, JSF, tcp/ip).
Not given enough time; at first we were supposed to finish development in like 7-8 months. But it was designed over several years. I kind of laughed, it was something like 90% design and 10% coding, yea right.
Couldnt retain key technical people; we had 3-5 head developers over like 6 months.
March 11th, 2007 4:47pm
I take offense at you for placing TCP/IP in the J2EE stack :)
I wanted to ask you what sort of mileage you were getting with all the newfangled J2EE stuff, anyway. I was googleing around and ran across you by coincidence recently looking for Spring in association with JRuby. I'm sort of involved with the complete redesign and architecture of a big webapp/enterprise thingie and wondering if using J2EE and especially Spring is worth the hassle. Pure Ruby is out of the question unfortunately and to be honest, I'm not much of a rails fan, either.
Spring is great at taking all the convoluted logging/transaction/database/configuration/loadbalancing and what have you out of your java code, but at the unfortunate expense of moving it all into way more complicated, convoluted xml files...
"I take offense at you for placing TCP/IP in the J2EE stack :) "
I don't know; I dont know of many projects that integrated spring and jruby but for what I was doing, it worked. You may have a speed issue because of jruby's pseudo interpreted nature.
On Spring, we didn't really use Spring in that project, but I would use Spring if I were to work with any J2EE. A lot lighter than any J2EE (except with maybe Wicket?).
If you do integrate jruby (probably mainly for the action controllers) I would interested to see if there is any performance hit.
March 11th, 2007 5:11pm
I was thinking more along the lines of using jruby to shortcut programming all the business logic, and maybe ActiveRecord for ORM.
"jruby to shortcut programming all the business logic, and maybe ActiveRecord for ORM."
Same thing, yea that is what I meant.
You can see what I did, here with full source;
Only thing, I didnt get to clean it up, so there is a lot of redundant code. Also, ruby scripts were loaded from file on each request (some caching could have been performed).
You might go with jruby's rails approach, maybe easier, I dont know.
March 11th, 2007 5:22pm
"Over-designed; business people had way too much control and screen layouts with billions of fields"
This isn't because the business people were in control; it's that the wrong business people were in control. Just like saying that "an app that nobody wants to use" is what happens when the developers are in control.
March 11th, 2007 5:30pm
I'd say that does tend to be quite usual behaviour for business people. Unless they have suitable background they tend to panic and think they have to make sure to include the kitchen sink, otherwise the door closes forever.
As project lead you have to try to calm those fears, and get everyone in the mindset of starting out with a simple, solid design, validating it, and building on that.
But if the company culture around you, and the people directly above you, don't support that idea, you're basically screwed. And the fact you ended up doing 2 years of design and 9 months coding says that something like that was going on.
From when you saw that happening though, you should have been putting your hand up and saying "this is not going to work".