Relation between site and host

[GoogleCodeExporter]

How to relate a site to a host? The relationship is mentioned in a couple of 
elucidations.

Original issue reported on code.google.com by alanruttenberg@gmail.com on 22 May 2012 at 3:57

[GoogleCodeExporter]

I've been assuming part_of (AKA 'is part of').

The full BFO doc, using anatomy as an example (and FMA as a precedent), makes 
the case for material entities having immaterial parts.  If I understand the 
doc correctly, those immaterial parts are Sites.

Original comment by dosu...@gmail.com on 22 May 2012 at 10:22

[GoogleCodeExporter]

There are two cases: Interior spaces - cavities, and exterior sites such as the 
example Barry gives of the environment of a cow feeding. (See attached image). 
In the latter case such sites can overlap others. So image a modern dairy farm, 
in which there are separate feeding pens for each cow. What, exactly, is the 
site part of? Suppose it is part of the pen. Then overlapping spaces and sites 
being part of their hosts would mean that the pens overlap, which would be 
counter-intuitive, particularly if the pens are independent artifacts.

In the FMA, I think that the cavities that are considered part of a material 
entity are so because there is a developmental program in which both are 
created. In the case of sites that are not created by developmental program - 
e.g. by fiat, or by the nature of nutrient exchange, or in order to control air 
traffic, I think the relation would be different.

Original comment by alanruttenberg@gmail.com on 22 May 2012 at 2:48

[GoogleCodeExporter]

Oops forgot attachment

Original comment by alanruttenberg@gmail.com on 22 May 2012 at 2:48

Attachments:

• Sites.PNG

[GoogleCodeExporter]

The document is already too long, and this is a hard problem to resolve. Hence 
I suggest keeping this at the level of elucidation, for BFO 2.0.

Original comment by ifo...@gmail.com on 22 May 2012 at 2:56

[GoogleCodeExporter]

"ELUCIDATION: a is a site means: a is a three-dimensional immaterial entity 
that is (partially or wholly) bounded by a material entity or a 
three-dimensional immaterial part thereof. [034-001]
EXAMPLES: a hole in the interior of a portion of cheese, a rabbit hole, the 
interior of your bedroom, the Grand Canyon, the Piazza San Marco, an air 
traffic control region defined in the airspace above an airport, the interior 
of a kangaroo pouch, your left nostril (a fiat part – the opening – of your 
left nasal cavity), the lumen of your gut, the hold of a ship, the cockpit of 
an aircraft, the interior of the trunk of your car, the interior of your 
refrigerator, the interior of your office, Manhattan Canyon)."

All of this seems clear enough - and I see no reason not to use part_of for all.

This is the first time I've heard anything about developmental processes in 
this context.  Not even sure this makes sense to me:  Would it discount a 
cavity formed by cavitation of a solid structure, but include rolling into a 
tube.  Even if it does make sense, it's rather impractical given the many cases 
where the precise nature of the developmental process involved will be unknown. 
 Perhaps you are misinterpreting the reference to 'Bauplan'?

As to the 'environment of a pasturing cow' - the main problem seems to me that 
more explanation of the intention behind the example is needed for us to have a 
sensible conversation about it. 

Original comment by dosu...@gmail.com on 22 May 2012 at 3:21

[GoogleCodeExporter]

"In the FMA, I think that the cavities that are considered part of a material 
entity are so because there is a developmental program in which both are 
created" - I agree with David that this is problematic. For example, sometimes 
a cavity is formed when the barrier between two pre-existing cavities ceases to 
exist - this may be an independent process from the development of the 
structures bounding the cavities.

I personally find lumens being part of MEs a little unusual (but this is maybe 
more of an FMA issue).

Original comment by cmung...@gmail.com on 22 May 2012 at 10:56

[GoogleCodeExporter]

A conservative point of view would avoid (yet again) overloading part_of (the 
new is_a). Let's call it as Varzi has, host_of and then (in another iteration) 
figure out the consequences for considering host_of a subproperty of has_part.

I don't like the FMA's using part_of either, and I have a nagging feeling, 
somewhat bolstered by discussion with Tom Bittner, that we will run in to 
problems using part_of. I'll be reassured by logical proof of consistency, not 
by one or the other you saying part_of seems fine. Part_of (finally now) comes 
with some axioms. Show me that the axioms don't lead to contradictions or 
counterintuitive results for seemingly simple cases (e.g. the two cows, for 
example, and a dozen more) and I'll feel more confident that part_of is a good 
choice. 

Note that Reference seems to agree with FMA when it wants, and not when it 
doesn't want to. So we agree about part_of for host/cavities, but deny their 
use of bounded_by to relate boundaries to what they bound (with bounded_by not 
a subproperty of has_part).

Original comment by alanruttenberg@gmail.com on 23 May 2012 at 1:50

[GoogleCodeExporter]

"For example, sometimes a cavity is formed when the barrier between two 
pre-existing cavities ceases to exist - this may be an independent process from 
the development of the structures bounding the cavities."

Case 1, the disintegration of the barrier is part of normal development: In 
this case the development process I mean includes that process. The barrier (as 
you define it) is part of the first set of structures bounding the cavities. 
It's ceasing to exist, if part of normal development, leads to a new structure 
that developed from the first. Not sure why you would call this independent 
(from a developmental point of view).

Case 2, the disappearance of the barrier is accidental. In this case we are 
talking about pathological anatomy, which FMA doesn't address.

In any case, I think the issue is that there seems to be a recent gravitational 
field operating on relations, pulling them over the event horizon into part_of. 
Part_of, already being somewhat inscrutable, gets harder to understand by 
virtue of these overloadings, and despite there being are axioms of mereology 
(arguably providing the only meaning part_of has)  never seem to actually be 
exercised to see whether extensions actually work.

As an engineering practice (and as a conservative ontological practice) it 
seems a better strategy to be safe than sorry.

Original comment by alanruttenberg@gmail.com on 23 May 2012 at 1:59

[GoogleCodeExporter]

Don't blame part_of, I don't think it's inscrutable or overloaded. On the 
contrary, once you pick your set of mereological axioms it wins the prize for 
the being the clearest, least problematic relation.

The inscrutability (at least for me) comes with BFO classes such as ME and IE, 
and how the axioms for these classes involving part_of are constructed. If IEs 
can be part_of MEs then I have no idea how to start determining where 
boundaries of MEs lie except without recourse to dodgy geometric 
simplifications. I think the problems you mention with cow pens above are just 
the beginning. I'd like to see some of Thom's theories.

But part_of is the innocent party here. The problem is with ME and IE axioms.

Original comment by cmung...@gmail.com on 23 May 2012 at 2:44

[GoogleCodeExporter]

The reason I blame part of is that the axioms relating to mereology are formal 
and not sufficiently tied to properties in the world (IMO). Any relation that 
satisfies the axioms is a fine part of relation, yet we always get into 
arguments about what is and isn't part. What I would prefer to see are part 
relations that are more clearly tied to physical phenomena - ones which can be 
objectively agreed-upon to hold, and this is why I was leading the direction of 
developmentally part of for the FMA discussion.

That said, I understand what you mean about them being the least problematic. 
They are, at least, supposed to work in a certain way, and the way it works has 
been put to useful purpose.

Bottom line, part_of is woefully underspecified - there are many many relations 
(considered extensionally) that can satisfy the axioms. On the other hand I 
think our intuitions about part of cover a relatively small set of those 
relations - we just don't say precisely which. For material entities our 
intuitions don't do *too* badly, but even there there are arguments that arise 
in anatomy. Mix in the IE and all bets are off, at least with what's been 
offered in terms of support thus far.

Original comment by alanruttenberg@gmail.com on 23 May 2012 at 2:59

[GoogleCodeExporter]

I agree that having IE be part of ME seems a bit odd, but (with the one 
exception below) I don't see any nasty side effects from this choice, so, in 
the interests of making some progress, would be happy to stick with it.  I was 
working from the assumption that we're translating the official BFO2 doc into 
OWL and this pretty clear states that part_of should apply between sites and 
their hosts. 

Having thought some more - I think that treating internal spaces as sites and 
therefore parts is fine.  I also think that the concept of the extended 
organism is potentially useful (see e.g.  Dawkins - The Extended Phenotype; 
http://en.wikipedia.org/wiki/The_Extended_Phenotype).  But treating various 
chunks of the environment of an organism as parts is asking for trouble.   
Given this, where are the limits of the organism?  Do we define them with 
(presumably) non-rigid criteria such as 'grazing range'?  If so, surely there 
are many overlapping and constantly changing external spaces that are part of 
an organism.

Original comment by dosu...@gmail.com on 23 May 2012 at 9:37

[GoogleCodeExporter]

Issue 12 has been merged into this issue.

Original comment by alanruttenberg@gmail.com on 23 May 2012 at 4:06

[GoogleCodeExporter]

Original comment by alanruttenberg@gmail.com on 23 May 2012 at 4:12

• Changed state: Accepted

[GoogleCodeExporter]

I suggest we check in sample/test ontologies into svn to properly explore the 
consequences of each design choice (this is a good idea anyway, a kind of 
"junit" test suite for bfo). I'd like to see how blood vessels/lumens/blood 
would be represented.

Extended organisms and phenotypes are useful concepts. There are a number of 
possible mereological sum combinations of "core organism" + microbes + internal 
spaces + various external spaces (e.g. OGMS uses convex hull, which is 
arbitrarily geometric). Do we need to materialize all as classes? The microbe 
issue is the most pertinent one for an anatomical ontology.

Original comment by cmung...@gmail.com on 23 May 2012 at 7:34

[GoogleCodeExporter]

+1 on tests.
The OWL2 working group created an (OWL) data structure for test cases as well 
as test harness software to run and report on them. See 
http://owl.semanticweb.org/page/Test_case_ontology 

The test cases are collected in a (semantic?) mediawiki. Perhaps someone is 
interested in trying to get a clone up and running for us?

Original comment by alanruttenberg@gmail.com on 23 May 2012 at 11:01

[GoogleCodeExporter]

[deleted comment]

[GoogleCodeExporter]

I didn't finish a sentence in the above, and didn't address the issue of 
host/site. So after posting this message correcting things I will delete the 
previous:

A discussion with Barry clarifies that a site occupant is located_in a site 
(continuant-continuant relation). The spatial region of the site is (will be) 
by definition big enough that the spatial region of the site occupant is part 
of it. 

Regarding the relation of host to site, part-of has been discussed, but 
boundary dependence might be a better relation. Barry indicated he would add 
boundary dependence to the reference.

Original comment by alanruttenberg@gmail.com on 28 Jan 2013 at 6:58

[GoogleCodeExporter]

If the relation is part of then histories must be restricted to material 
entities, as otherwise the 1:1 relation continuant to process fails due to the 
possibility of two continuants occupying the same spatial region (e.g.) the 
site of a puddle and the water filling it. 

Note also that in a discussion with Werner that two continuants could be 
located in the same place was a surprise and he thought it ought to be 
modified. 

My preference would be to have a distinct relation at least, and possibly a 
distinct relation from sites to regions to avoid the distinct continuants 
having identical regions.  

Original comment by alanruttenberg@gmail.com on 9 May 2013 at 2:12

• Added labels: Milestone-bfo2-release