Re: [Axiom-developer] Re: [sage-devel] Re: Identification with ldap.

[root]

>> Do you want your code to live?
>
>That's a very interesting question. When I was much younger, I used to
>think I did. But when you realize that an awful lot of code isn't meant
>to "live" -- it's meant to solve an immediate problem, and then be
>discarded -- then you don't care whether it lives or not.
>
>Axiom was clearly designed to live, rather than solve an immediate
>problem. But how much code is out there that was designed with a 30-year
>horizon? 

More to the point, how much code is in Sage that is designed for the
30 year horizon? Will a matrix inversion still give the right answer
30 years from now? If so, how many dozens of times will it have to
be reimplemented because nobody can understand the current code?
Which of the 16 kinds of matrix inversions are used and how do they
differ in terms of input domains, performance, robustness, etc?
Should I use the Maxima inversion or the Sympy inversion or the
Lapack inversion or implement a new one? 

The front ends will likely change. Axiom's hyperdoc help system is
being reimplemented using a firefox front end. The build system is
about to be moved from using 'make' to using 'asdf'. The file system
layout is disappearing, replaced by latex literate books. 
<http://axiom.axiom-developer.org/axiom-website/documentation.html>
Sage's notebook is certain to be replaced.

Even so, if we want the next generation of developers to contribute
effectively it seems important to write down what they need to know so
they don't give up in frustration. That's a serious loss. We need a
computational mathematician to be able to contribute effectively
without becoming frustrated. We also need to capture the special
knowledge that lies behind the contributed code since there will be
very few experts worldwide in a given research area. Would you 
recognize and be able to fix a bug in a feynman quantum computation?

William has stated that he doesn't care about this issue since he
wants a system NOW that he can use to do his number theory research.

But it is fundamentally about the quality of the system. The 4Ms
don't care because they pay people to learn the code.

Google and Microsoft, having invested real dollars into the
effort, ought to try for a higher standard and a longer horizon.
I really wish they would hire a team with a mandate to make this
code literate.

Tim