We are working on a private svn repository: this dir is only a copy,
so please don't touch the files, otherwise it will be a nightmare to
keep the two versions in sync.

Suggestions, remarks and corrections are very welcome: please write
them in this file.

--------------------------------------------------------------------

[fijal]
* re-using native frameworks is not the main reason for lack of performance
  of CLI vs Psyco. It's also because there is no notion of dynamic/virtual
  structures, ie there is no way to tell that W_Int is becoming int
  just-in-time
* I think we mostly use meta-programming instead of bootstrapping (also
  I don't think you can write boot-strapping, but niko is proper english
  speaker, not me)
* "The standard implementation of Python, even though reasonably fast,
  is based on a VM which is not so portable and robust as the CLI or the JVM."
  - this is simply not true. It's actually more portable. Basically, if
  you assume "JVM on every platform", why not "Python on every platform"?
  There is a gap something in this reasoning.

* I would suggest using notation from D05.1, instead of source-code SomeXxx.
  Sounds at least silly. Also it reads that SomeList provides some methods.
  While this is not true, SomeList does provide an annotation for list methods.

* I failed to find that all globals are considered constant.

* mixins - there should be also example of a function returning a class
  with memo. So it will just work as expected, just need to be static-enough.

* overall - I am not a good reviewer for that. I (somehow) fail to see the
  point, also because I fail to understand RPython as a dynamic language
  and I don't like Java and C#, so convincing me is easy and hard.

[/fijal]