One thing that occurred to me the other day, are specific types still necessary or a legacy that is holding us back. What I mean is: do we really need short, int, long, bigint etc etc.

I understand the reasoning, variables/objects are kept in memory, memory needs to be allocated and therefore we need to know how big a variable can be. But really, shouldn't a modern programming language be able to handle "adaptive types", ie, if something is only ever allocated in the shortint range it uses fewer bytes, and if something is suddenly allocated a very big number the memory is allocated accordinly for that particular instance.

Float, real and double's are a bit trickier since the type depends on what precision you need. Strings should however be able to take upp less memory in many instances (in .Net) where mostly ascii is used buth strings always take up double the memory because of unicode encoding.

One argument for specific types might be that it's part of the specification, ie for example a variable should not be able to be bigger than a certain value so we set it to shortint. But why not have type constraints instead? It would be much more flexible and powerful to be able to set permissible ranges and values on variables (and properties).

I realize the immense problem in revamping the type architecture since it's so tightly integrated with underlying hardware and things like serialization might become tricky indeed. But from a programming perspective it should be great no?