The father of C + + Bjarne on STL templates in C + +

In 1994, my main concern was how to make the ISO C + + standard as good as possible--while in the two aspects of the characteristics and specifications it contained--and get the majority's consent. Even if people do not accept a certain specification, it will not affect the good of it (normative). The ISO standard is not mandatory, so some people think that they are not worth wasting time adapting to it unless the pressure of the [group] community can convince them of the value of the specification. Adapting to the environment is an important additional task for an implementing person, so adapting to the environment is a conscious decision and requires allocating resources that could have been used elsewhere. Some obscure language features are difficult to implement in some compilers. We can implement or buy a class library, and the leading, reliable implementation (implementer) has the opportunity to "lock" users with their imaginative proprietary features. So, I think the point is: let the members of the Committee and the organizations they represent believe that the standard document is the best document they expect to see.

After doing a lot of work, the Committee was successful. At the Morristown (New Jersey,usa) meeting in October 1997, the final vote of the technical members was 43-0. We had a celebration after we learned the result! In 1998, the ISO Member States approved the standard with an unprecedented 22-0 vote. In order to gain the unanimous consent of the Commission, the Committee has done a great deal of technical work and has used a number of diplomatic strategies: At that time, I like to say that "political problems cannot be solved; We must find the technical problem that raises the problem and solve it". I can't imagine just voting, because the minority obeys the majority to simply "solve" the problem, at the same time, because "the political bargain" the question also endangers our best technical judgment--and this question (template's separate compilation) still is "deteriorating", needs to look for a better technical solution.

In the year preceding the final vote, the Committee's work was:

1. Details, details and more details.

2. STL

3. Separate compilation of templates