Re: [Bug-apl] Question about GNU APL

[Elias Mårtenson]

On 19 February 2015 at 21:28, Blake McBride <address@hidden> wrote:

On Thu, Feb 19, 2015 at 7:10 AM, Juergen Sauermann <address@hidden> wrote:

 

1. those things are probably not very hard to implement. However:

2. I try to minimize non-standard extensions of GNU APL because every such extension
    creates incompatibilities of APL programs that are using them. My idea of free software
    is that not only GNU APL itself should be free but also APL programs running on GNU APL.
    And for a free APL program to be useful it is important that is is portable between different APL interpreters.

 

I agree with Juergen.  The structures you suggest (although I would add a 'while') make a lot of sense, but so would a lot of other things. There are many single-source, non-standard, hodgepodge languages out there that few are interested in.  APL is APL.  It is what it is, and it has proven itself valuable as it is.  GNU APL's full support of APL2, along with its keyed file system, make GNU APL a very powerful and proven standard.  Adding a bunch of hodgepodge extensions would ruin it.

Just one opinion.

Christian is probably not aware of the numerous emails that has covered this on this mailing list in the past. I myself raised this very early.

I fully respect Jürgens opinion, and I certainly have no intention to try to go off in a different direction.

With that being said, I think it's well known here that my opinion in this matter is diametrically opposite to that of Blake. I see GNU APL being one of many GNU-titled programming language implementation projects, and in those (for example, the GCC project), the direction is quite clearly different, and one that I mostly subscribe to.

The idea is that extensions originating from other platforms should not be opposed, but rather welcomed since it has many benefits. For one, it makes the software offering more appealing to users of commercial platforms, thus helping the GNU project as a whole. Secondly, adding extensions that are unique to one's own platform help drive general progress forward, by making the system more useful, and at the same time give users a reason to stay with the system.

Finally, a technical note (that I believe I've made before). I think Jürgen said at one time that he felt that the Dyalog-style :If, :Then, :Else structure is ugly, and I most definitely agree with that. However, I feel that GNU APL is seriously lacking flexible iteration constructs. Either some kind of LOOP structure, or support for functional programming so that looks can be implemented using Scheme or Haskell-style.

Regards,

Elias