Re: Temporal validitity of subject indicators?

Hi Robert,

On 9 May 2011, at 9:52 AM, Robert Cerny wrote:

> Andrew S. Townley wrote:
> 
>> Hypothetically, let's say you're defining a topic for "Bob's Garage" and you know that Bob's been in business for 20 years with the same telephone number.  A perfectly good choice for a subject indicator might be his website, http://bobsgarage.com or even his telephone number (via RFC2806), tel:+19135551234.  You know that these might have the added advantage of appearing near any other references to Bob's Garage so you pick one or both.
>> 
>> Fast forward to time t+n, and Bob's goes out of business, the hosting company redirectsbobsgarage.com to a link farm/advertising page and Bob's phone number gets reassigned to something else.  The problem is that your topic map (authors and users) don't really have any idea that this has happened, so they get new data with these references, but the references now indicate different subjects.  At time t+n+1, someone discovers that the subject indicator is no longer valid and needs to update the map.  You might also just happen to have some old data around that needs to either be referenced or used to rebuild the map.  Some of that "old data" may also be in the form of serialized topic maps.
>> 
>> What do you do?
> 
> You realize that 'http://bobsgarage.com' was *not* a perfectly good choice for a subject indicator.

I don't agree with this.  I'm not saying that it's a subject indicator that meets the criteria of a *published* subject indicator, but, in the absence of anything else, I don't see why as a general subject indicator, the website owned, managed and updated by the subject themselves isn't a "perfectly good choice" for an SI.

If you introduce an intermediary, they're always going to be relying on the source for keeping track of things like this.  In the dynamic world of business, people and other ordinary things in the real world that haven't achieved the global or local fame to have someone dedicate time and effort into maintaining a "fan page" about them on some controlled site or something like Wikipedia.

Subject identifiers of any kind are varying degrees of arbitrary and have varying degrees of stability--even those maintained by an individual topic map author within a closed system.  This is just a function of how systems and the real world work.  Even with the best original or initial intentions, they're never going to be guaranteed to be valid beyond a certain time interval around the point at which they're assigned to a topic.  Thinking otherwise, or worse, requiring otherwise, is just going to get you into trouble eventually or seriously limits the scope and applicability of any given map.

I still fundamentally believe that identity assertions at any particular point in time are relatively arbitrary and *extremely* context-dependent.

> Generally, your idea reminds me of an open space session that Lutz Maicher and Xùan Baldauf were giving at one of the TMRA conferences (2008?). They suggested to scope subject identifiers. I have never seen a similar emotional reaction to an open space session and it was not positive. There was a follow up discussion on this list, if i remember correctly. This shows you that this is an important and delicate subject which touches the heart of the community. I thought it is a good idea, since i became aware of the semiotic problems that embrace any semantic technology during my work on Topincs. The following resources helped me to understand the subject matter better:
> 
> * Sebeok, Thomas: Communication Measures to Bridge Ten Millenia
> * http://www.topincs.com/people/robertcerny/758
> * http://www.topincs.com/people/robertcerny/1036

I must've missed this.  I've only managed to make TMRA '09, and I didn't remember the discussion on the list.  Thanks for the references, though, and I'll see what I can learn from searching the archives and looking at the resources you mention above.

> On the other hand i think that the TMRM does it right, by leaving the actual source of information untouched and by putting the algorithms to reach subject identity decisions into legends. Actually when developing my idea of content and glue topic maps i looked at the TMRM and wondered why it does not have this problem. My answer was that it actually lets people use the identifiers they are most comfortably with (their own) and delays the burden of deciding on subject identity to the last possible moment (late merging). This is what content and glue topic maps is about.


I remember our extended discussion about your concept of glue maps at TMRA '09.  I can understand where you're coming from with the idea, but I don't think the concept scales well.  If you're in a tightly controlled environment with a small authoring team, I think it might work as you intend.  However, if you publish a merged content+glue map somewhere, that too becomes part of the ecosystem, and it seems that the model breaks down since you've lost the control and distinctions of content vs. glue maps consumed by other people.

I don't read the bit about "leaving the actual source of information untouched" in the TMRM, but that's certainly a valid interpretation.  All it really says (to me) is that there isn't a fixed notion of what constitutes subject identity as part of the model and that the manner of resolving whether two subject proxies represent the same subject is defined in application-specific terms within a given TMRM legend.  Whether this identity resolution requires or facilitates merging of proxies is also application-dependent, as the purpose of needing to resolve proxy identity is also not covered.

Cheers,

ast
--
Andrew S. Townley <[email protected]>
http://atownley.org