... So the question of whether names are "*the* (only) way" to obtain sharing isn't really a language question-- it's more of a compiler question.
Are they the only way that's guaranteed to result in sharing, or is even that not the case?
Depends on what you mean by "guaranteed". Since in Haskell sharing vs. not sharing is not a semantic issue, you can't very well say that if sharing doesn't occur under such-and-such circumstances, the language specification is violated.
But the answers to the original question seemed to be saying you would get sharing if you wrote code of a certain sort. I take it that the language semantics doesn't strictly require sharing, but it sounded like it was something programmers could assume would happen in practice. If not, then shouldn't the answers to the original question have been different? What I was wondering was why the answers to the original question about pointers involved giving a name to the thing you wanted shared. Is there something less direct that would also work? For instance, could I instead name something that contained the structure I wanted shared and then use an appropriate expression to refer to the part I wanted shared - or do I have to give a name directly to that part. The main point of my Java example was to give an example where I didn't give a name to the very thing I wanted shared, but instead gave a name to something that contained it.
This is very different from languages like Java, where
Object[] x = new Object[](...) Object[] y = x;
means that x and y refer to the same array (not merely equivalent arrays), and if they don't, it's not Java. And what's crucial is that you can write a Java program that detects whether x and y refer to the same array, by updating the array via x and observing the effect via y.
You can just use ==. Also, I'm not sure the Java language semantics really does require that x and y in your example above refer to the same locations in memory. The update test doesn't necessarily show that either. It just shows that if you change one, the "other" changes too. Of course, programmers assume that assignment to array elements is done in a fairly direct and simple way; but the language semantics might nonetheless allow something more expensive and complex. -- Jeff