[WikiEN-l] The disambiguation virus

Tue Feb 4 23:49:00 UTC 2003

Hi,

the short version: Using the format e.g. "Java programming language" for  
disambiguation is bad, because it leads to more links of the same style,  
even where no disambiguation is needed. Since we have the "pipe trick", we  
do not need this natural "disambiguation" at all.

The long version:

I've observed an effect which I call the disambiguation virus. As its name  
suggests, it is a harmful effect that decreases the coherence of the wiki  
by making linking harder.

The disambiguation virus requires another idea as its host: That "natural  
disambiguation" is better than disambiguation in parentheses. An example:

        "Java programming language"

        vs.

        "Java (programming language)".

It is easier, this idea suggests, to link to [[Java programming language]]  
in the text than to [[Java (programming language)]].

However, in many cases, this is not true. For example, when I write

"C++ has a clear speed advantage when compared with higher level languages  
like [[Java programming language|Java]]."

I have to write more than when I write

"C++ has a clear speed advantage when compared with higher level languages  
like [[Java (programming language)|]]."

What you see here is the "pipe trick", which removes rightmost text in  
parentheses from the displayed link title.

But that idea is not the actual virus. The virus is far more dangerous.  
Once someone starts a page with this "natural disambiguation", others  
quickly add to it. I have already moved dozens of pages from "foo  
programming language" to "foo". These pages did not require any  
disambiguation! "COBOL" is obviously better as a page title than "COBOL  
programming language".

The same is true for web browsers (so we get "MidasWWW web browser"),  
languages ("Hamburger language", which is also inaccurate) and cheeses.  
Once someone has started, other people use this format for all of the  
links on the same page, whether they require disambiguation or not.

This is not the case with the "foo (bar)" format, which looks somewhat  
uglier and is therefore avoided. But the "foo bar" format reads natural,  
almost like we should always have this kind of qualifier in titles. After  
all, it's nicer to know what the page is about by reading the title,  
right? Wrong! Most pages that link to the title have already established  
the context, there's no need for us to repeat that. This is evident from  
e.g.

http://www.wikipedia.org/wiki/Alphabetical_list_of_programming_languages

This page actually uses only links of the form

[[foo programming language|foo]]

ARGH! Where context is really needed, it is better provided in the text  
than in the link.

In conclusion, I suggest we avoid the "natural" format and move pages to  
the "unnatural" one to reduce infection rate.

Regards,

Erik