Re: [Gzz] Linked or associated nodes vs. xuLinked enfilades

[Benja Fallenstein]

(Trying to understand what you're saying. Yell when I misunderstood.)

Matti Katila wrote:
> Shortly thinking xuLink might be a good answer for linking problem(in 
> structure level you can't link more than one node to one(at least 
> currently (this is not well enough explained!, fyi))). By saing shortly I 
> mean in time.

"At first, xuLinks may seem like a good solution to the linking problem.
However,"

(I do *not* understand the part in parantheses, by the way.)

> For example you have canvasA where is an annotation what is 
> linked to B,C and D (xuLink).

"suppose you have a canvasA containing a note which is linked to notes 
B, C and D (through a xu link)."

> You copy the annotation text to another 
> canvasB but all xuLink of course follow.

"Suppose further that you copy the text in the note to another canvasB. 
Of course, the copy is still linked to B, C, and D."

> So, now you are not interested 
> about, let's say C and D.

"Maybe you are not interested in C and D in the context of canvasB; 
i.e., you want to keep the link to B, but not the links to C and D."

> How can I unconnect those but keep annotation 
> in canvasA to be right?

"How can I remove the link to C and D from the text in canvasB, but 
still have it in the text in canvasA?"

> Even association is in canvasA to B, C and D that doesn't implication
> that association with that text should be in canvasB with C and D or even 
> B.

"Even if the text in canvasA is linked to B, C, and D, that doesn't 
imply that the text in canvasB should be linked to C and D-- or even B."

> With node structure we can change that but with xuLinks we can't, and 
> that implications that xuLinking can't be used.

"When we represent associations as RDF connections instead of xu links, 
we can connect the RDF node on canvasA to C and D, but not make the same 
connection from the RDF node on canvasB. With xu links, this is not 
possible. Therefore, we cannot use xu links."


Welcome to the Xanadu tradeoff.

The tradeoff here is: Say that you have the same text (same characters) 
in two different places. Are links to the *characters in a specific 
place* or to *the characters wherever they are*?

If the links are to the characters *wherever they are*, the characters 
may be copied into a different document where the link makes less sense. 
For example, a paragraph from a play may be linked to an analysis of the 
role of this paragraph for the whole play. When a sentence from the 
paragraph is quoted in another document, it will still be linked to the 
analysis-- even though linking the single sentence to the analysis of 
the *paragraph*'s role in the *play* may make little sense.

If the links are to the characters *in a specific place*, links may not 
be shown in another context where they are also relevant. For example, 
if you link a word from a paragraph to a dictionary definition, if this 
paragraph is also on another canvas, on the other canvas the word will 
*not* be linked to the dictionary definition.

This is essentially unsolvable: When making a link, the computer cannot 
decide on which of the transclusions this link makes sense, and on which 
it doesn't. The human making the link doesn't have the time to make the 
decision for every instance (especially when there are many different 
versions of the link's target around).

However, sorting out the noise ('this link is out of context here') is 
much easier than guessing where an interesting link may be ('maybe 
someone has made a link to one of the 144 versions/transclusions of this 
word? Let's look through them all!').

Therefore, Xu chooses to make links to characters *wherever they are*. 
This is not perfect; but in an unavoidable tradeoff, it's the better choice.

Can you give an example where it's a big problem if a link goes wherever 
some text is copied? It seems to me that the problems there are much 
smaller than the other way around.

- Benja