| | 1 | = Adding regression testing of top-100 websites with invariant analysis = |
| | 2 | |
| | 3 | Layout tests could not be extended with real online websites such as top-100 (top-300) lists due to the changing content. However, it would be possible to extract a set of invariant attributes for these sites, such as basic layout, number of links, etc, and use these attributes as a base for comparison to expected attributes. |
| | 4 | |
| | 5 | The invariants need to be determined first by hand and then encoded in a unified description as a testing oracle which is suitable for automatic processing. The actual invariant attributes could be determined based on rendering information and this way the comparison could be performed completely automatically. |
| | 6 | |
| | 7 | The manual analyis of top-100 websites have been performed and now we are working on formalizing the attribute set. See technicalities at [wiki:"Top100InvariantDetails"]. |
