Subject: Re: FSBs and mechanized documentation
From: Rich Morin <rdm@cfcl.com>
Date: Mon, 6 Mar 2006 11:06:31 -0800

Quinn Weaver has made some interesting comments off-list.  I'm
posting them here, with my responses, to aid the discussion.


At 12:20 PM -0800 3/5/06, Quinn Weaver wrote:
> What about a search engine that just works with the content
> of the system?  If I could do the equivalent of a Google mail
> search on my bug-tracker-plus-source-repository, it would make
> me really happy.  Is this what you're getting at?

Well, search engines are very useful, but they are also limited.  I
use Wikipedia for some things, rather than Google, because I want
an organized response.  In general, if an organized and relevant
information source is available, it will be more convenient than a
grab bag of keyword "hits".  In point of fact, the Semantic Web is
partially motivated by this observation.

Here are some questions which a search engine couldn't handle, but
which an integrated system could:

  *  What percentage of the bug reports were logged during the
     development, testing, and post-release phases?

  *  What components (e.g., modules, functions, files) are "hot
     spots", getting far more than the median number of changes,
     bug reports, etc?

  *  Having found a document which describes a particular aspect
     of the system, how do I direct a question or comment to all
     concerned parties?

  *  Which files (e.g., in Linux) are connected to a particular
     (e.g., CUPS, gcc) subsystem?


At 10:30 AM -0800 3/6/06, Quinn Weaver wrote:
> By the way, if you're going to build this thing, I hope you'll
> build it as a framework with pluggable back-ends.  For instance,
> many people will want a tool that works with Subversion, but I'd
> like one that works with monotone, my source-control system of
> choice.

Any integrated system for mechanized harvesting of information is
going to need a variety of "input filters".  Sometimes these can
build upon report-generation capabilities of the subsystem being
monitored.  Other times, it may be easier to dive straight into
the data.  In any case, the system needs to have a way to fold in
fairly arbitrary information about entities and relationships of
interest.

-r
-- 
http://www.cfcl.com/rdm            Rich Morin
http://www.cfcl.com/rdm/resume     rdm@cfcl.com
http://www.cfcl.com/rdm/weblog     +1650-873-7841

Technical editing and writing, programming, and web development