One of the arguments for keeping APCS-A as is are the costs to change. There are of course, many kinds of costs.
The first are the monetary costs. New text book, curricular materials and possibly equipment. The College Board has their development costs but I don't care about those - they're raking in money hand over fist anyway. I don't think these costs amount to much. Companies that sell materials for classes are always trying to make sales and even when subjects don't change there are always new books and other resources. So, these costs shouldn't really figure into switching languages.
Another cost relates to ancillary effects. Does changing the language or changing the course have an effect on how well or how poorly students are prepared for what's to come. The argument here for keeping Java is that CS2, data structures in most colleges is taught in Java or C++ so switching APCS-A to Python wouldn't prepare students as effectively. This may or may not be the case. Students already have to deal with the switch from Java to C++ if they go to a C++ school adn get placement out of CS1 and some colleges do indeed teach CS1 in Python and then switch.
So, changing languages won't necessarily hamper students that take data structures in a different language in college.
On the other hand, if the course content changes significantly - changing from the A in APCS-A meaning Algorithm to API as Owen Astrachan suggested at SIGCSE then that's a different kettle of fish.
Digging a little deeper on this front is that when a college designs their CS0, CS1, and CS2 classes they should look at the full sequence - all three of these classes and beyond. If you change a course, you really have to think about how it changes the future ones. Hopefully colleges do this when changing curriculum but I did talk about how Hunter didn't when it rebooted their CS0 class. Now, this really isn't a reason to change or not to change since a high school can't really track and pay attention to every colleges CS0, CS1, CS2 sequence so I guess it's neither here nor there.
I guess the bottom line for this cost is that as long as APCS-A remains fundamentally the same content wise it may not matter what the language is and if the College Board clearly states that APCS-A is not going to be a college CS1 then any language is fair game.
The biggest cost that people bring up is teacher training. All the teachers will have to learn the new language and possibly a new paradigm. This one is huge. I work with a lot of teachers who are relatively new to CS and some of them are struggling their way through APCS-A. These teachers are basically students in CS0 or maybe CS1 and just finding there way to and through CS. Switching languages for them is no easy feat and concepts in one language don't necessarily translate to another. I've seen teachers who claim proficiency with Python or JS struggle to write the simplest of Java programs or after spending a significant amount of time in Java they have similar struggles switching to Python.
In spite of so many people saying CS is easy and anyone can do it, learning new things is in fact hard and takes both time and effort.
As a side note, this cost flies in the face of every PD provider (I'm looking at you code.org) that claims "we've trained XXX teachers to be CS teachers in just YYY sessions!"
This is the state we're in today but this is also a warning. We have to do better with respect to teacher training so as to make this cost meaningless.
When I entered teaching, I had been doing C programming professionally but knew Pascal along with other languages from college. Teaching APCS-A/B in Pascal was easy. Well, the language was easy. I was a new teacher so teaching overall was hard but teaching APCS-A/B wasn't any harder than say teaching Trig or Linear Algebra. Actually, maybe it was a little easier since I was a CS guy at heart and not a math person.
When the class switched to C++, I figured out C++. I had never done OOP before but I figured it out. Switching to Java? Same deal. I also taught myself Python, NetLogo and other languages that I had to or wanted to teach with.
The point is, I had a strong CS background so picking up a new language wasn't really a major cost. It wasn't free - I had to spend some time and at times figure out new tooling but it wasn't this insurmountable thing.
Right now, we're at a point where we can't really ask this of our CS teachers. Too many are too new but we have to make sure that our pre-service training programs and state requirements are such that in the future we will have CS teachers that know enough CS so that a language switch in a course isn't a big deal. Rather, a teacher will be able to do an analysis of the new tool solely based on asking if it's right for their students.
So, those are the costs. Are they enough of a reason to keep Java? I don't think so. If you give enough lead time, as pre-service programs are getting up to speed you can minimize the training costs. Just give it 10 years instead of 5. Of course, you have to be confident that Python will be the language that far into the future.
That said, I'm not advocating for a switch to Python. We still have to go back to my first post on this - The College Board has to decide if this will be a college CS1 class or something else and only then can we do the language vs language analysis.
Are we done with this thread? Almost but I think it's worth spending some time looking at some possible alternatives to both Java and Python.
That's next time.