Hi, Tom Brinkman wrote: <comments about the introduction> I felt exactly the same after reading the introduction. Then I figured it might be about "one-shot allocation" and reading the rest of this thread it seems that assumption has been correct. Anyway, I still don't understand why we can't have intrusive containers without intrusive syntax. E.g: struct slist_node // <-- part of container implementation { T element; // note: held by value slist_node * next; }; Then there's only one allocation per 'slist_node'. With compiler options that enable structure padding (up to 2^n, for alignment) it will be bad, but usually it can be disabled for individual structures (in this case one would have to disable alignment/padding for whatever T is). So, why does the user have to add custom data members? Regards, Tobias