You wouldn't believe how good a relationship with a client will be if they understand ehat you do ;)
http://forums.construx.com/blogs/stevemcc/archive/2007/11/01...
Short-term technical debt can speed development, but like credit card debt, it quickly accrues interest. If you let your technical debt accumulate, then eventually you're spending all of your energy servicing the debt instead of improving your software.
The hardest part would be having some way to quantify the debt. But you could, at least, make a record of the corners cut, and make any accumulation extremely visible to management. Does anyone know of anything like this?
We could have a pretty accurate debt precision in time, that time could be used to fix things ?
My guess is that you either don't have PMs that say no or there is a really poorly structured chain of communication (ie - you have clients contacting more than just their PM for changes) which is changing the PM/Client relationship.
It is largely about drawing a line in the sand. If you let your clients overrun you with requests, then they will keep doing it. Why wouldn't they? Get billed the same for more features? Sign me up. Luckily I have PMs watching out for me.
The reality is that the "fine line" is the more reasonable and correct approach. It's case by case, every time. If there was a black-and-white rulebook, Joel Spolsky, Jeff Atwood, or Patrick Mackenzie would have blogged about it by now.
Not educating clients is the worst thing we can do as software professionals.
Why do most people not ask for a house with a rushed foundation or frame? Because most people know that that's asking for serious issues later.
Why do most clients ask for key software for their business rushed out the door and delivered ASAP, sans testing and without proper spec? Because most don't know that they're asking for serious issues later.
But when a client makes a request like this, often unaware of what may actually be easy or fast or simple or consistent, I can't say no to it. Saying that we need to take more time to write tests and document it properly would likely come across as similarly unprofessional and reflect badly on the company. And the consequences are every bit as bad as you say they are: the code rots and becomes volatile and obnoxious to maintain.
That's why it hurts to read this post--it feels so painfully true. I'm lucky in that at my company, my project managers are coders themselves who understand the programmatic constraints and the danger of software rot--but my clients usually aren't. I have no solution for this; it's a fine line to walk.