Regarding SSO, it's a case of whether 2 out of 3 is good enough. I
draw your attention to the middle S, sequence. Documentation of the names for
the packages, classes, methods and signatures are sequence independent. This
is to be contrasted with many programs where the 2000th instruction has to be
executed before the 2001st, otherwise a pointer to null or invalid memory
is
dereferenced and the execution halts prematurely. I've got a
personal
java
package called /myname/.util. If I gave you the documentation, you could
replicate regardless of the order of the specific methods in my or any
listings.
There is one place where sequence does matter: within the
signatures
of methods that take 2 or more homogeneous parameters. For
instance, a
method that looks like this: Object o getObjectFromInts(int i1, int
i2), the
output could depend on whether a value is in the first or second
position. In
order to implement the method, so that it replicates how someone
else
implemented it, I need to see their documentation. As a practical
concern, the method name of a well designed library will hint to the
programmer the actual order of parameters. After all, a language shouldn't ask
a programmer to look up the documentation every single time a method is
used. I take it that SSO is not a case where two out of three ain't bad.
Well, maybe it ain't bad, but I suspect it isn't good enough. [ Reply to This | Parent | # ]
|