The advantage is that, supposing the a pointer to the head of the list is stored in some other container, like struct Container { struct Foo *first

The advantage is that, supposing the a pointer to the head of the list is stored in some other container, like struct Container { struct Foo *first_foo; ...probably more stuff... }; then the first entry can have its `prev` field simply be `&container->first_foo`. (Each subsequent entry, as the OP said, has its `prev` field be `&previous_entry->next`.) That way, you can have a function that removes an entry from the list, given only a pointer to the entry: void foo_remove(struct Foo *foo) { *foo->prev = foo->next; if (foo->next != NULL) foo->next->prev = foo->prev; } No need to keep separate track of which container you belong to, and no need to special case the first entry of the list. The FreeBSD sys/queue.h macros use this pattern for the LIST type. (The disadvantage of this scheme is that despite it being a mostly doubly linked list, you can only iterate forwards.)