This is the mail archive of the guile@cygnus.com mailing list for the guile project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

RE: New manual plan


Reynolds, Gregg writes:

 > I suggest functional organization with lots of indices and xref lists;
 > use XML if you want to avoid duplicate data; and most importantly, drop
 > the Grand Plan of The Manual approach in favor of a pamphlet approach
 > (or: prefer the bazar to the cathedral) - make it the Guile
 > Documentation Library.  Establish editorial policies, an XML DTD, some
 > categories, and maybe guidelines on how to classify a pamphlet and write
 > a synopsis for inclusion in the master index, etc.  I think it would be
 > much easier then to get others to write doc about their favorite little
 > corner of Guile, and you wouldn't need to worry to much about
 > integrating things.
 > 
 > I'd especially like to see a detailed "How Guile Stacks Up to
 > (perl/python/vb/you-name-it)" pamphlet or series of pamphlets.  I could
 > have used one a few weeks ago when my boss was looking at Scheme v.
 > Python as an extension language.  

Pamphlets are much more amenable to machine-production.  I support this
approach.  Like all pamphlets, the value is in whether or not point of
view is enhanced through simple association.  Views are important, as
the UI people have discovered.  What is documentation but static UI?

A view has value when it is consistent and, to some point, goal oriented.
View selection algorithms must take this into account.  Consistency
should not be completely exact, but allow for some distribution in search
results.  Views can be registered (PITA) or inferred.  The inference
should be through direct use (this is the feedback loop).  I vaguely
recall some (US) east-coast project doing personalized-info strategies.
And there are many indexing solutions being researched.  Would it be a
good idea to develop a guile module for that stuff?

Anyway, w/ views, documentation (and indeed, what is everything but
documentation) can be dynamic, and if there are good big-brother type
packages, :->, one could even turn off one's brain completely and have an
agent (hey, gush mailing list is quiet!) serve it up as you hack.  Of
course, I think paperclips are tacky.

As an aside -- I wonder if forum2000.org is GPL?  My guess is no, since
there are commercial interests involved.  At,

	http://www.cs.cmu.edu/~andrej/quadratic/.html

3.2 point 2 algorithms should be GPL, IMO.  The last question in the
hall of fame doesn't make me hopeful.

Anyway, the point of my rambling is: dynamic objects are nifty.

thi