Some views on software development

Automated testing takes a relatively small cost, and the processing logic of the unit being tested must be separated from the data it processes, since both the program and the test code need data. Those in the C environment of the programming master, Unix source code are relatively good to do this.

A good software must be better than the implementation of user needs, otherwise, no matter how delicate your design, in the user's opinion, it is not worth much. For example, data maintenance software can be a lot of input into bulk input, specify a rule, the user according to this rule only enter common things once, and then enter the various resources in different places, and then at the time of the import point of error, this use with the user a record, each record may also error, Users are absolutely happy to use them.

Software Use error found problem responsible person is very difficult, in the interface of a business processing error, interface developers find backstage, but the background of the program log is not complete, background developers spent three days, check data, tune code, finally found another subsystem processing errors, Perhaps the developers of this subsystem will take a long time to discover whether it is their own problem or a third-party problem.
From the management, the essence of this problem is that the developer and maintenance personnel are not separated, or the developer will certainly write the log better, it will be easier to find the wrong subsystem.
From the software design, this is the principle of separation of duties is not observed, network communication and business processing is a different responsibility, if separated, it can be more convenient to record the network communication messages sent and received, there will be no such problem.