这是一个消极的结果，作为一个“基础创新”很奇怪，但我认为适用，因为它开辟了新的研究领域，关闭了无用的领域。

分配共识的不可能性:2001年PODC影响力论文奖

We assumed that the main value of our impossibility result was to close off unproductive lines of research on trying to find fault-tolerant consensus algorithms. But much to our surprise, it opened up entirely new lines of research. There has been analysis of exactly what assumptions about the distributed system model are needed for the impossibility proof. Many related distributed problems to which the proof also applies have been found, together with seemingly similar problems which do have solutions. Eventually a long line of research developed in which primitives were classified based on their ability to implement wait-free fault-tolerant consensus.

不确定1980年的情况，但人工智能社区几十年来一直是一个创意生成器，他们仍然在做。

我没有资格在一般意义上回答这个问题，但仅限于计算机编程?并不多。

为什么?我思考这个问题已经有一段时间了，我认为我们缺少两样东西:历史感和客观评价我们所创造的一切的方法。并非所有情况都是这样，但大体上是这样。

For history, I think it's just something not emphasized enough in popular writing or computer science programs. Take language features, for example. A canonical source might be HOPL, but it's definitely not common knowledge among programmers to be able to mark the point in time or in which language a feature like GC or closures first appeared. And of course after that there's knowledge of progression over time: how has OOP changed since Simula? Compare and contrast our sense of history with that of other fields like maybe political science or philosophy.

至于判断，这确实是我们寻求成功的客观衡量标准的失败。给定foobar，它以什么可衡量的方式改进了编程行为中的某些方面，其中foobar是任何设计模式，敏捷方法，TDD等等。我们有没有试过测量这个?我们到底想测量什么?正确性，程序员的生产力，代码的易读性等等?如何?软件工程确实应该着手解决这些问题，但我还没有看到。

我相信单元测试、TDD和持续集成是1980年之后的重大发明。

采用面向对象。

这个想法早前就出现了(例如Simula)，但在20世纪90年代成为主流。(恕我直言，它最大的好处之一是在开发人员之间提供了一个通用的词汇表，所以它的广泛采用使它更有价值。)

在操作系统核心开发中使用函数式编程/语言。