Volker Wysk wrote:
> I think, using literate programming techniques could be very useful. One
I'm not sure about this, particularly in Haskell.
There is a world full of literate programming tools for VLLLs
(very low level languages), where they should be much more
necessary than in Haskell, which is rather human-readable
by itself.
Experience seems to indicate that their usage ist rather
limited. SGMLing the fortrash sources at netlib would be far
more helpful than well-written code in whatever VHLL.
In addition, there are some technical problems with such methods:
- hackers prefer ergonomic editors, which essentially limits
them to vi or emacs (protecting ones hands from RSI is more
relevant than beautiful code presentation).
- having the <foo bar="baz"> stuff *visible* would make the
code unreadable, but AFAIK there are no tools for emacs to
hide them (or, better, interpret them as fontification
hints for the content)
(I recently learned in comp.text.sgml that psgml is still
considered the best SGML editor available).
> Maintaining code and documentation at the same central place makes it much
> easier to keep the docs accurate and up to date. Editing several files
> each time one does a change, would be tedious.
I'm not sure about this either: Most changes are bugfixes and don't
need to be mentioned in the manuals, but only in the changelogs.
Additionally, people who are good at hacking code are not always
those who can write nice user docs. An index of identifiers might
help in some situations, but IIRC most identifiers in FP languages
are either local (so mentioning them would not make sense) or have
long, expressive names.
Given the fact that humans are lazy, we would probably end up
with embedded LP-docs contradicting the actual code instead of
contradicting separate files (and the latter can be fixed by
a helpful soul without messing with the source file itself (which
would cause unnessecary version changes in RCS etc.)).
Ralf
