[erlang-questions] FOP (was: Re: Trace-Driven Development)

[Jan Burse]

Michael Turner schrieb:
> The usability of a*given*  wiki? Or the fitness-for-purpose of wiki
> software for helping maintain Erlang/OTP documentation, from an
> administrative point of view? It's only the latter issues that concern
> me here, and it's on those issues that you keep saying things I find
> either incomprehensible or flatly wrong.

There is an further issues with client-server based
collaborative tools such as a wiki. You need to have
all the time a working internet connection. This is
ok if you are doing work from a location with a stable
and fast connection.

But sometimes the check-out and check-in model is
superiour to the client-server based model. Especially
if you have intervalls where you don't need a lot of
interaction with peers and/or where connectivity is not
possible.

Since all your artefacts are local after check-out
you will also experience a much faster responsiveness
of your tools. Since one is allowed a smoother editing
one usually has a higher productivity. Also before
check-in again changes are often automatically
compressed.

There are a couple of integrated development environments
either free or for a few bucks that integrate well with
version management, and that also provide graphical diff
tools. These diff tools can be used for many things like
merge conflict resolution, change reviews, change promotion,
etc.. Here is a typical diff view:

http://codingbone.files.wordpress.com/2010/11/picture-2.png

I am currently involved in changing a wiki solution
with a Prolog driven automated kind of proof reading
into a check-in and check-out solution. Most probably
the outcome will be a hybrid, but in the future the real
editing will be done kind of offline. And the proof reading
load will be done locally and not anymore on the server,
resulting in much more speed.

But prerequisit for all this is also a kind of locality
respectively predicatabilty of the involved data, i.e.
pages. Low frequency higher grain check-out is only
possible if one can predict what module one will need
for a particular task at hand. In a memex like environment
where one is about to follow unpredictably links, this
is difficult.

On the other hand in a hybrid solution, the wiki could
still be used to browse accidential links. And I guess
in a structured work environment it will be also possible
to delineate modules. But I am not yet sure how well
things will work. Problem are always highly shared modules
which have a lot of updates. Check-out and check-in can
also slow down an organization. But good automated
merging might help here.

Bye