The use-mention distinction

["Drummond Reed" <drummond.reed@cordance.net>]

Title: Re: [xdi] on global cross references and the link contract use case

> Les wrote:

>

> What is the english difference between these two statements:
>
> +international+employment$contract$sig$d
>
> And

>
> +international*(+employment)*($contract)*($sig)*($d)
>

 

With the caveat that English equivalents of XDI RDF statements are only an analogy, here’s the translation (assuming $sig = “signature” and $d = “date”):

 

            XRI                   +international+employment$contract$sig$d

            English             international employment contract signature date

 

            XRI                   +international*(+employment)*($contract)*($sig)*($d)

            English             international “employment” “contract” “signature” “date”

 

The key point being that an XRI cross-reference expresses roughly the same concept as the English concept of the Use-Mention Distinction:

 

            http://en.wikipedia.org/wiki/Use_mention_distinction

 

As explained in the introduction to that Wikipedia page:

The use–mention distinction (sometimes referred to as the words-as-words distinction) is the distinction between using a word (or phrase) and mentioning it. For example, the following two sentences illustrate use and mention of the word cheese:

• Cheese is derived from milk.
• Cheese is derived from a word in Old English.

The first sentence is a statement about the substance cheese. It uses the word cheese to describe its referent. The second is a statement about the word cheese. It mentions the word without using it.

In written language, mentioned words or phrases often appear between quotation marks ("Chicago" contains three vowels) or in italics (When I refer to honey, I mean the sweet stuff that bees make), and some authorities insist that mentioned words or phrases must always be made visually distinct in this manner. Used words or phrases (much more common than mentioned ones) do not bear any typographic distinction.

**************

The irony behind the http://wiki.oasis-open.org/xri/XriThree/GcsDelimiter proposal is highlighted by the last sentence above where it says, “Used words or phrases (_much more common_ than mentioned ones)…” (emphasis added).

 

Our history in XRI is that first we recognized, way back in the prehistory days (2000) that XRI syntax needed a way to encapsulated (and thereby reuse) identifiers from other contexts (most specifically, absolute URIs, but also relative URIs and other string-based identifiers). We invented the parenthetical cross-reference syntax to do that. Since that was the ONLY mechanism we had for reusing identifiers, we also used it to reuse global XRIs in the context of other global XRIs. That meant that the ONLY way to refer to =drummond in the context of @cordance was to express it as either:

 

            @cordance*(=drummond)

            @cordance!(=drummond)

 

Therefore, to use the terminology of the Wikipedia articile, we semantically interpreted such a cross-reference as a _use_ of =drummond and not a _mention_ of =drummond.

 

However as we began to run into issues with restricting global XRI reuse to cross-reference syntax, the idea first arose that GCS characters could function as delimiters just like LCS characters (* and !). That meant we could express one global XRI directly in the context of another global XRI, e.g.:

 

            @cordance=drummond

 

This syntax was directly parallel to English, where the direct _use_ of one English word in the context of another English word requires no special syntax other than creating an ordered set of the words. This is expressed in the last sentence of the Wikipedia quote above:

 

Used words or phrases (much more common than mentioned ones) do not bear any typographic distinction.

 

From an XRI standpoint, suddenly everything snapped into place. The ordinary _use_ of one global XRI in the context of another XRI could be accomplished simply by creating an ordered set of the global XRIs, without any special syntax. This would align it directly with English. Then the special exception, i.e, the _mention_ of a global XRI without intending its normal use, would be represented by cross-reference syntax. That means in English:

 

            XRI                   @cordance=drummond

            English             Cordance Drummond

 

            XRI                   @cordance*(=drummond)

            English             Cordance “Drummond”

 

Following the use-mention distinction, the presumption would be that the =drummond in @cordance=drummond is the regular _use_ of =drummond, and therefore is a reference – in Cordance’s context – to the same entity referred to by =drummond in its own (global) context.

 

By contrast, the presumption would be that the =drummond in @cordance*(=drummond) is a _mention_ of =drummond, and therefore should NOT be interpreted as a reference to =drummond in ordinary _use_. (What it should be a reference to is up to @cordance as the authority for this subsegment.)

 

Personally, I think it is a combination of: A) the evolutionary history of XRI, and B) the subtlety of the use-mention distinction, that has take us so long to recognize and understand this issue, which is why I don’t blame Les for really pressing us to explain it.

 

=Drummond