Providing mappings can be quite a bit of extra work for a project. What languages do they provide mappings for? Python, Vala, Java? Then that project needs developers who actually use the target language regularly, otherwise they are likely to provide a poor mapping that doesn’t integrate well with the language’s standard API. They also need experience in the target language to test their mappings.

On the other hand, if the mapping is done by the people designing the language, they can ensure that the resulting API integrates well and makes full use of the language’s abilities. But you’re right: they may not be interested (or even know about) a certain library then leading to API skew.

From the developer’s point of view, I generally prefer an API that is provided by the language because it then means I have access to all the documentation in the same place. It also gives me more confidence that the mapping was done carefully and that using that API won’t result in ugly memory leaks or other issues.

To be pragmatic, I think the key is to find out who benefits more in having the mapping available, and therefore who is more willing to dedicate resources in writing, testing and supporting the code. In the case of gudev and Vala, I would say that Vala has more to gain from the mapping because it adds an essential API to the language. In the case of a specialised library having, say, a Python mapping, the project will have more interest in writing the mapping because it will then make their library available to a very large group of developers.