Ian Woollard wrote:
Well, let's take an example, like:
[snip]
We're talking about a bit different things. You're trying to convince
me that in practice ILLs often represent "similarity" (BTW, it's a good
example). But I already know this.
What I'm saying is that there's a big discrepancy between the policies
and the reality.
And at no stage is the linkage strictly wrong. The
underlying problem
is that you're assuming exact correspondence, whereas it's more like a
thesaurus; these are *synonymous links*. The phrase for people in the
know is: 'There's no such thing as a true synonym.' And that's what
blows it up.
It's Wikipedia: cities, countries, famous people and dates are examples
of articles that either have the exact correspondence in another
editions, or no correspondence at all.
Thus, if you start from [[:en:Griffin, Georgia]] and end up in
[[:en:Classic RISC pipeline]], it makes more sense to correct some of
the *incorrect* ILLs that caused it, rather than conclude: well, that's
what ILLs are, let's get used to it.
The problems are many fold. Linked articles can have a
definition that
makes them a subset, partial overlap or superset. If you go through a
few rounds of going to subset and then partial overlap and back up to
superset you can end up practically anywhere, as you've shown rather
admirably. There is absolutely no reason to think that these links are
transitive in practice or theory.
Some ILLs represent subset/superset, others represent loose similarity,
others are simply wrong and yet another represent exact correspondence
(yes, there are such ILLs!) Wouldn't it be nice to know which are
which? Right now part of the community interprets all the ILLs as loose
similarity, while another part interprets all of them as exact
correspondence.
Regards,
Ćukasz
PS. OpenPGP somehow ate all this, leaving only the last paragraph which
already came in the previous post. Sorry for that.