Effectively implement any basic strategy, usually need to go all out. - "Competitive Strategy"
Following the first chapter to propose the five-way model, Potter proposes three basic strategies in the second chapter - the leading strategy of total cost, Differentiation, target agglomeration strategy [1].
Didn't think about what the concept on software development can correspond to the cost, first don't consider the first strategy. The latter two strategies are very understanding, in Chinese, it is "another way to take a different path" and "the operation of the industry."
Take me, don't understand anything when you just entered the company, people are hard-working, I will not help. Let's take a look at the document, read the textbook, and the opportunity to be bored. The company wants to support the debugging code. This part of the working theoretical is not strong, most is a trivial technical details, just right for me to become familiar with people. Go up. It can be said that the optimized code is extremely difficult to debug, and there is no forming solution, so it is not high for our requirements, first put the frame, which functions can be done first, the strongest requirement is that the debug option is not Impacts the generation of optimization code, and the generated code must be exactly the same as if the test option is not added. In this way, under the tolerance of everyone, I will read the textbooks to understand the basic theory. When the task write code understands the specific implementation, I can complete the analysis and implementation of the independently after half a year, I started to rewrite the entire support with one year later. Debugged frame. Although it is limited to the function of supporting, the effect is not ideal, but for me, it is too much. It is not exaggerated, now I can completely compete for the task of optimization. In fact, I have modified a module algorithm (not related to debugging support), this module is considered to be one of the most complicated modules in our compiler. .
This is another way to start from optimization, but start from debugging. In fact, this is a company in order to cultivate new employees, and only in other people's tolerance, there is a certain limitations, it is not easy to imitate. However, when I cross the first door, I can choose the development route with my ability and experience. Continue to do debugging, or turn to do optimization is a good choice, but I have an idea, I plan to make another way, but this time I can't expect others to be tolerant, I must first convince others, I have to do it. How important, then make things out on the ground. If you can successfully leave a "said a different approach person", then take a lot of resistance, it will be a lot of resistance.
Let's talk about "the operation is specialized". In addition to being robbed to modify the module algorithm once, I have been doing the work of debugging optimization code, and the front will last for more time. As the function is increasing, the situation has undergone subtle changes. Debug can not be destroyed, but optimization does not destroy the resulting debugging results, and also specialize in a set of test sets for this. If some colleagues are not interested in support debugging, or if they don't have time to understand, they will invite people like me to check their work (REVIEW), ensuring that they don't destroy what. Of course, the test set is the last level. If the test is not, we must find us to help debug. In this way, I set up a prestige in a relatively narrow field, establishing a barrier, thereby reducing competitive threats.
Potter mentioned at the end of this chapter, any strategy is risky, details will not be described here, summarizing eight words - trial time, with the times.
[1] To tell the truth, these three words have a little awkward.
