Re: Discussion on the structure and purpose of /info/dir

[Eli Zaretskii]

> Date: Tue, 05 Mar 2002 01:03:02 -0500
> From: "Peter J. Farley III" <address@hidden>
> 
> Aha!  Now I see why we disagree so much on this subject.  We have 
> exactly opposite methods of searching for information.  My method of 
> searching is to start at the most general first, if necessary weeding 
> through large swaths of useless chaff to decide how to get more 
> specific in my search criteria.

This is very inefficient, in my experience, at least when using the
index search.

> Your recommendation of trying 
> "--apropos cut" first is exactly the reverse of how my mind 
> works.

I didn't propose "--apropos cut", I proposed to use these, in the
order indicated:

      --apropos 'cutting text'
      --apropos cutting
      --apropos cut

This is, of course, schematic, but the point is to try the specific
string first, then gradually make it more general until you get some
hits.

> Plus, knowing the name of the utility "cut" is exactly what my 
> example searcher did *not* know.

"cut" is not the name of a utility, it's a verb here.

> When I am searching a book (a reference book), I begin with the Table 
> of Contents, looking for chapter headings that might relate to the 
> information I need, then scanning down sub-chapter headings for more 
> detailed information, and then finally reading actual pages in the 
> subchapters to find the actual information.  I don't generally start at 
> the index in the back, since I usually don't know any of the detailed 
> specifics of the subject at hand.

I never look at the TOC except if the index search fails (or there's
no index).

The way to search the index is to find a name for the issue you are
looking for, then look up that issue in the index.

The indexing in GNU manuals is designed for such searches.

> Using web search engines is a very similar experience.  One starts with 
> a general term, then adds additive and subtractive search terms to make 
> the search more and more specific, until there are few enough entries 
> to review in one sitting.

I use the Web searches the same as I described above: start with
several words which I require to be present, then drop words if I
don't get any hits, until I get some.

> That is the kind of structure and content I would like to see in 
> "/info/dir", that would allow searchers to start at the most general 
> and work their way down to the most specific.

IMHO, it won't work because the collection of GNU manuals lacks
hierarchy, which is necessary for the kind of top-down search you are
used to.