The strict separation of the operators from the algorithm and the ability to easily add operators, has its roots in rather social considerations.
Normal coding algorithms are limited to the encoding methods of one or a few ideas of some (in the order of 1 to 100) people (the development team). But there are world wide much more people (on the order of probably 100000) who deal in a broader sense with efficiently and easy encoding of multimedia objects, thereby inevitably many ideas to code multimedia objects remain ignored. Even if some of these people have an interested in contribute their ideas, this is very difficult or impossible to realize.
Everyone can contribute new ideas to the Fib algorithm in the form of new operators (a term for this is "crowdsourcing"). To do this, it is sufficient to only possess the knowledge for the Fib multimedia description language and the interface for the operators of the algorithm. In order to introduce new ideas or operators, for efficiently coding of multimedia objects, no adjustments to the algorithm or other operators should be required. This will limit side effects and any implementation of an idea has only to worry about the idea respectively its operator. So there are no further knowledge needed of the algorithm or even other operators. In this way the system can grow, without increasing the complexity of the (base) system and with this the difficulty for maintenance and expansion of the system.
The algorithm and the image description language should be designed so that layman can train themself without much effort to implement new ideas respectively operators. In particular for students in computer science (or people with computer science as their hobby), it should be possible to train themself within a few days, so that they can implement their own operators and integrate them. (How useful or effective this is, should be of no concern here.)
The GNU GPL license, under which the algorithm is published, clarifies the legal situation for new operators. This allows to add new operators to the algorithm, unless other rights are infringed (an injury would be, for example, that the operators use an algorithm or code which are under incompatible licenses). New operators can also be made available to the public.
The rating of the operators should motivate people to integrate their own operators. Thus, anyone who has added an operator can realisticly assess, how good his operator is in relation to other operators in a situation. Competition among operator authors should be beneficial to the creation of new and better operators.
All this should result in that not only a small development team will contribut to improve the encoding of Fib objects, but that a much larger group of people are concerned with this issue. This should give the development and propagation of Fib a further boost.
In this sense, the genetic algorithm for Fib is a trans ingenious algorithm, that is designed to extract the ideas of people to encode multimedia data from their heads and then transport/collected them into one pool. Thus, these ideas/algorithms can accomplish more than they could individually. The algorithm can thus combine more intelligence than one human (or a small group) can produce.