"always passes by reference"

[Ulrich Schreiner]

you try to describe highlevel features with lowlevel arguments. this
will fail. i'm currently working with linux/KDE to write this answer. i
see a window with a grey background on the screen, a toolbar and much
more. in some way these things are 0s and 1s in the memory of my
graphiccard but you won't get happy with this binary view. 

in my opinion it is the same with computer languages. when working with
high level oo languages you really do NOT work with 0s and 1s, you are
working with objects. when programming with c or assembler you have to
know about addresses and memory locations and much more. otherwise you
will fail. but you can not compare this c/assembler view with an
objectoriented view. you will fail too if you try to describe highlevel
language paradigmas with the assembler view (every language will
implement it in another way).


"(Greg Weeks)" wrote:
> So you phase it.  I would never talk that way myself, and if someone asked
> me to describe what was going on, I would use quite different words, eg:
> Each object is the address of a region of memory -- an address that can be
> written to multiple locations.  You prefer to think of the object as a
> region of memory [I think], and you define a writing of that memory's
> address to be a "reference".  I prefer to think of the object as an
> address, and I don't give any special name to a writing of that address.
> 
> There is no right or wrong here, but there is also no single correct
> description of what is going on.  Some say Python is call-by-reference,
> some say Python is call-by-value.  Everyone is correct.
> 
> Greg
> 
> PS: My motivation for my particular phasing is that I like to view all data
> in the computer as patterns of 0s and 1s.  (I find this tidy and easy to
> remember.)  An address is a pattern of 0s and 1s.  A block of memory is
> not.  So, for me, a datum can be an address, but not a block of memory.