Next: Issues to check for
Up: Correctness and Performance
Previous: Profiling
  Contents
  Index
A standard technique to check for consistency in the development is a reviewing process. Each module of an application goes through a review process, where persons not connected with the development check the deliverables (source code and documentation) for completeness and consistency. This review process serves multiple purposes:
- It forces the developer to finish a version of the program with a certain polish.
- It helps to find inconsistencies or missing explanations in the documentation.
- It encourages ``best practice'' in the ECLiPSe application development, bringing together experts from different application teams.
- It helps spread knowledge about applications and their sub-systems, so that re-use opportunities are recognized earlier.
On the other hand, a number of problems are normally not recognized by a review:
- The review checks one version of an application at a given time point. It does not guarantee that changes and modifications after the review are performed to the same standard.
- A successful code review does not imply that the application code is correct. Reviewers might sometimes note suspect code, but a review cannot replace proper testing.
- If nobody actually checks the code, then the whole process becomes useless overhead. This means that resources must be properly allocated to the review, it is not a task that reviewers can undertake in their spare time.
- Comments and change requests in the review must be recorded and acted on. A formal review comment form may be used, alternatively we might work with detailed and complete minutes.
Next: Issues to check for
Up: Correctness and Performance
Previous: Profiling
  Contents
  Index
Warwick Harvey
2004-08-07