Definition of Project Success

I have recently read an article from CIO about project management, I strongly agreed to Tom Klein's comment: "Projects don't fail, the expectations do". Yes, unless something is purely digital (only "1" or "0"), otherwise, most of our measurements are "relative" basis and "the expectation" does play an important role of defining "success" or "failure".

So how do we define a project success?
I used to work in a company when it ran with a series of death-march network infrastructure projects (high risk, under resources, tight timelines.....). All of them were finally completed (actually, there was no chance for even a failure because of the nature of the projects were related to the fundamental IT system infrastructure of the company affecting the entire business operation [the bread & butter]). But I would not say those projects were successful indeed.

The hidden failure in a project success
Because the "death-march" project nature, those projects had to be executed using the "dotcom mode" (the mode of business development/operation during the dotcom era around 1996-2000) where everything proceeded in ad-hoc and agile basis without thorough planning, well-defined process, completed documentation, sufficient testing, and etc.
Eventually, those projects were completed, but affiliated with lots of negative side results: several production incidences during the project implementation stage affecting the business operations, demolishing the company's well established project management process,  damaging staff morale, high HR turnover and dismantled trust among teams (conflicts and politics in resources/priorities arrangement against other well planned projects).

In theory, the successful criteria of a project should be measured by its deliverable against original project objective in terms of time, scope, quality and realized business benefits. However, within a business entity, a project's success should also be measured in a holistic (program management) view where the indirect, intangible benefits (or loss) from the project should also be counted.