I have recently become irritated by having to code dummy arguments for routines to be used as agents. Eric Bezault suggests that Eiffel could be usefully extended to eliminate the need for this.

So I have sketched a proposal here

Many people have said that they dislike having to export a routine just so that it can be used as an agent in a pre-condition. The objection is that if the routine does not naturally belong to the ADT, then it should not be exported to ANY. Well, this is a valid point, although there are already routines in ANY that do not belong in the ADT of most classes. Such as generating_type.

Following on from http://www.eiffelroom.com/blog/colin_adams/correctness_conditions_for_calling_an_agent, I have thought about the following scheme (it would require ECMA to be amended to require a tag on assertions, but this should be done anyway. Also, the tag would have to be unique with all the preconditions for a given feature - likewise for postconditions.):

<eiffel> class ROUTINE

I was recently reading the paper "From Design Patterns to Reusable Components: The Factory Library by Karine Arnout and Bertrand Meyer" and I noticed that although they had succeeded in turning the pattern "Abstract Factory" into a component, the resulting class (ABSTRACT_FACTORY) did not have a sound contract.