ontolog-forum
[Top] [All Lists]

Re: [ontolog-forum] Re: Semantics

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Internet Business Logic <ibl@xxxxxxxx>
Date: Mon, 02 May 2005 18:36:41 -0400
Message-id: <4276AB79.2060006@xxxxxxxx>
Hi All --

Frank McCabe wrote...

One definition of semantics is in terms of the objects/concepts of  interest and their relationships. I.e., to model the semantics of  some *thing* identify the objects of interest and then their  relationships.

I tried recently to make a partial order (lattice) of some of the things that people mean when they use the word 'semantics', such as:  logical model theory, real world URI types for data a la RDF, OWL ontologies, the meaning of an English statement**, etc.  Making such a lattice turns out to be a hard task.

But what one can do is to pull back to just those notions of 'semantics' that are likely to be useful in IT in the next few years.  I'd propose:

(1) declarative semantics of rule systems (model theory or fixpoint theory) and proofs of soundness and completeness of rule engines wrt to the theory

(2) In the spirit of Frank's relationships between objects of interest , some better way than we have at the moment of computing with relations between English sentences.

There's more about this in   http://www.w3.org/2004/12/rules-ws/paper/19  , and in the other papers in last weeks' W3C Rules Interoperability Workshop.

So, anyone disagree with (1) and (2) ?     There are surely some other notions of 'semantics' that folks on this forum think will likely be useful in IT in the next few years.

                                                        Cheers,  -- Adrian

** My favorite gripe about ontologies is that the only English sentences they contain are comments, which are simply ignored by machines.
-- 

Internet Business Logic  --  online at www.reengineeringllc.com

Reengineering LLC,  PO Box 1412,  Bristol,  CT 06011-1412,  USA

Phone 860 583 9677     Mobile 860 830 2085     Fax 860 314 1029



Duane Nickull wrote:
This thread just started on the SOA Reference Model TC list.  Any comments?

Duane


-------- Original Message --------
Subject:     Re: [soa-rm] Semantics (was: Re: [soa-rm] Re: Autonomous Services?)
Date:     Mon, 02 May 2005 13:26:49 -0700
From:     Francis McCabe <fgm@xxxxxxxxxxxxxxx>
To:     Duane Nickull <dnickull@xxxxxxxxx>
CC:     soa-rm@xxxxxxxxxxxxxxxxxxxx
References: <955953BEFC7F5C4AA0F1428F3F335F1CE327B8@xxxxxxxxxxxxxxxxxxxxxxxxxx> <4274B7B4.4080208@xxxxxxxxxxxxxxxxxxx> <427655BD.4050704@xxxxxxxxx> <7FD249D5-A9A6-477A-98B6-D091BA15006E@xxxxxxxxxxxxxxx> <42766875.60606@xxxxxxxxx>



Ahh, well, you begin to point at something deep here... about the  relationship between syntax and semantics.

One definition of semantics is in terms of the objects/concepts of  interest and their relationships. I.e., to model the semantics of  some *thing* identify the objects of interest and then their  relationships.

Syntax, on the other hand, is all about written _expression_; and the  manipulation thereof.

The problem that most logicians skirt over is this: in order to  describe the semantics of some *thing*; you have to write it down --  i.e., you need syntax to do semantics. Something of a paradox.

I have come to the conclusion that the relationship (sic) between  syntax and semantics (proof theory and model theory), is primarily  one of perspective: semantics is *about* relationships, and syntax is  *about* structure.

Which is a long-winded way of saying that, yes, you can do a lot with  syntax manipulation. However, where simple-minded syntax manipulation  breaks down is in the broader context: I can re-write one form of a  telephone number to another because I (i.e., not part of any spec.)  happen to know they are equivalent for my purposes.

The other big issue with purely syntactic operations is that there is  no support for distinguishing valid transformations from invalid  ones. Often to prove a syntactic transformation is valid (sanctioned  by the semantics) is non-trivial. What a sound logic gives you is a  once-and-for-all proof that certain transformations are valid.

Frank

On May 2, 2005, at 10:50 AM, Duane Nickull wrote:

Frank:

I just skimmed this work and find it very interesting.  There are a  couple of sections we may wish to ponder in our context.

Section 1.1 - definitions.  Clean separation of the Semantic Model  (probably what also may be termed ontology) from the concept of  semantics itself.  There is also another separation of semantics  annotation (where something like UDEF would be categorized) from  the notion itself.

Section 1.2 - a great depiction of how semantics relate to Data Model.

page 15-16 - I do not see how this XSL instance to OWL mapping:

<xsl:template match="/">
<Address rdf:ID="Address1">
<has_Receiver rdf:datatype="xs:string">
<xsl:value-of select="POAddress/recepientInstName"/>
<has_StreetAddress rdf:datatype="xs:string">
<xsl:value-of select="concat(POAddress/streetAddr1,POAddress/ streetAddr2)"/>
</has_StreetAddress >
<has_City rdf:datatype="xs:string">
<xsl:value-of select="POAddress/city"/>
</has_City>
<has_State rdf:datatype="xs:string">
<xsl:value-of select="POAddress/state"/>
</has_State>
<has_ZipCode rdf:datatype="xs:string">
<xsl:value-of select="POAddress/zipCode"/>
</has_ZipCode>
<has_Country rdf:datatype="xs:string">
<xsl:value-of select="POAddress/country"/>
</has_Country>
</Address>
</xsl:template>

varies from what a schema can tell you.  The basic premise of  containership structure allows the same declarations IMO.

I did like the aspect of context specific semantic reference  however. The aspect of context is missing from many semantics  works.   An example - you define something called "party".  Seems  easy - a party is an entity assuming a role within an exchange (or  something similar to that).  The gotcha is in the implementation.   The semantics of "Party" varies from itself within different  hierarchic manifestations.

//PO/Buyer/Party != //PO/Seller/Party

This is potentially a bad example since it could be solved with  ambiguous references but the main notion is that every element has  at least one context qualified.  Luckily, we do not have to address  a solution for this but I hope these folks can.  It looks like  their xpath statement to OWL grammar works conceptually.

Duane


Francis McCabe wrote:


+1
In fact, I am hard put to understand how you can *store* semantics.
You can only store data. The best that you can do is store a   description of the semantics; but that is not the same thing.

On that theme, IBM and others at the U of Georgia recently  released a  paper on semantic annotations of Web services. Have  not yet had the  time to digest this properly, but could be  interesting... if IBM  makes a play in the standards space with this.

The link to the paper is:

http://www.alphaworks.ibm.com/g/g.nsf/img/semanticsdocs/$file/  wssemantic_annotation.pdf

Frank


On May 2, 2005, at 9:30 AM, Duane Nickull wrote:


John
(aka "Meggan".  Hey - how you dress in private is none of our   business  ;-)

Just joking!!

This is a good question.

The registry is one way that one could store semantics however   semantics are not required to be explicit and there are other   models for sharing information beside registry.  At the abstract   level it represents a facet of the model where the information   available is meaningful.  Therefore, a registry will not be in  the  reference model as a normative, core element.

We decided to add a non normative section to explain some of  these  manifestations.  How one goes from "Data Model" to  Messages,  Availability to Registry, Policy to on the wire  security etc.

It would be great if you could hook up with the person with this   section and offer proof reading services.  Value your input.

Duane




meggan hardin wrote:



My assumptions (so far) about the central metadata concepts  have  been that the reg/rep holds this data. Are we delving to  the level  of defining specific types of resources / components  that should  be included in a major component such as the reg/ rep? I think that  the concept of storing semantic metadata as  an independent  integration reference point is important enough  to be included in  the RM.

FWIW - Contivo terms the semantic metadata repository the   "enterprise vocabulary"...

john

Smith, Martin wrote:



Violent agreement.
 martin
________________________________

From: Schuldt, Ron L [mailto:ron.l.schuldt@xxxxxxxx]
Sent: Fri 4/29/2005 6:39 PM
To: Smith, Martin; Sharma, Sameer; Duane Nickull;   john@xxxxxxxxxxxxxxxxxxx
Cc: ebSOA OASIS TC; soa-rm@xxxxxxxxxxxxxxxxxxxx
Subject: RE: [soa-rm] Re: Autonomous Services?



Sameer will correct me if I'm wrong, but I believe that his   intent was
to include the notion of central metadata within a "Reference
Architecture" not the Reference Model. Appendix B is the place  where
example use cases would be defined. I suspect that Sameer might be
willing to submit an example use case.

Ron Schuldt
Senior Staff Systems Architect
Lockheed Martin Enterprise Information Systems
11757 W. Ken Caryl Ave.
#F521 Mail Point DC5694
Littleton, CO 80127
303-977-1414
ron.l.schuldt@xxxxxxxx


-----Original Message-----
From: Smith, Martin [mailto:Martin.Smith@xxxxxxx]
Sent: Friday, April 29, 2005 4:19 PM
To: Sharma, Sameer; Duane Nickull; john@xxxxxxxxxxxxxxxxxxx
Cc: ebSOA OASIS TC; soa-rm@xxxxxxxxxxxxxxxxxxxx
Subject: RE: [soa-rm] Re: Autonomous Services?


Sameer - -

Let me practice being Matt <g>:

The term " 'central' metadata" presumes a specific implementation
strategy and should not be in the RM.  Perhaps "metadata   associated with
the service should be available in the environment."  Now in  my  example
SOA for Appendix B, I'll probably show a UDDI services  directory, or
maybe a combo registry/repository that can in fact store all the
description metadata.

Martin



-----Original Message-----
From: Sharma, Sameer [mailto:sameer.sharma@xxxxxxxx]
Sent: Friday, April 29, 2005 2:16 PM
To: Smith, Martin; Duane Nickull; john@xxxxxxxxxxxxxxxxxxx
Cc: ebSOA OASIS TC; soa-rm@xxxxxxxxxxxxxxxxxxxx
Subject: RE: [soa-rm] Re: Autonomous Services?


My feeling is that some of what you are alluding to might be   covered by
UDDI,
however as is happening in an instance of SOA deployment that I am
involved
in - UDDI by itself is not going to be sufficient to express  all the
metadata
that is needed for a client to successfully and contextually   interpret
all
that a Web Service does.

My attempted solution is to capture this additional metadata by
leveraging
central metadata services of my enterprise. I guess what I am   saying is
that
the concept of "central metadata" might be a valid candidate as a
component of
the Reference Architecture we are considering.

Since I was unable to participate in the F2F, (due to some  last  minute
commitments that I got called into), if this topic was  discussed,  please
accept my apologies for bringing it up again.

Thanks!



L
  Sameer Sharma
    Principal Applications Architect
    Lockheed Martin Corporation
    Chief Technology Office (CTO)
    12506 Lake Underhill Road - MP 166
    Orlando, FL-32825
    Tel: (407) 306 5640
    Fax:(407) 306 1392

-----Original Message-----
From: Smith, Martin [mailto:Martin.Smith@xxxxxxx]
Sent: Friday, April 29, 2005 1:38 PM
To: Duane Nickull; john@xxxxxxxxxxxxxxxxxxx
Cc: ebSOA OASIS TC; soa-rm@xxxxxxxxxxxxxxxxxxxx
Subject: RE: [soa-rm] Re: Autonomous Services?

Folks - -

On my way home from N'awlins Wed night, I had a thought on this
discussion.

I think we expect services in an SOA to be independent of the   kind of
shared contextual knowledge we usually presume within a local   computing
environment. We expect that the requesting service will be able to
obtain all the info it needs to use the responding service   successfully
by processing the responding service's description metadata.   I  do think
this is a core characteristic of SOA services.

I'm not suggesting we reinstate the use of the word  "autonomous"  as a
handle for this concept since it demonstrably caused confusion  at  the
f2f.  If we need a handle, perhaps "self-sufficient" or
"self-documenting" or "introspective" (naaah - forget that one.)

Martin




-----Original Message-----
From: Duane Nickull [mailto:dnickull@xxxxxxxxx]
Sent: Friday, April 29, 2005 12:43 PM
To: john@xxxxxxxxxxxxxxxxxxx
Cc: ebSOA OASIS TC; soa-rm@xxxxxxxxxxxxxxxxxxxx
Subject: [soa-rm] Re: Autonomous Services?

We discussed and the submitter withdrew the submission pending
clarification on exactly what is meant by Autonomous nature WRT
services.  It may be re-submitted and probably will however we  do  not
have consensus on it at present.

Duane

john c hardin wrote:




Duane and SOA-RM group -
Can someone enlighten the members of the eb-soa group regarding a
description of Autonomous Services? Any resulting  conversations  from
the meetings this week, on the subject of Autonomous Services   would be







good also.

thanks
john





-- 
***********
Senior Standards Strategist - Adobe Systems, Inc. - http://  www.adobe.com
Vice Chair - UN/CEFACT Bureau Plenary - http://www.unece.org/ cefact/
Adobe Enterprise Developer Resources  -
http://www.adobe.com/enterprise/developer/main.html
***********












-- 
***********
Senior Standards Strategist - Adobe Systems, Inc. - http://  www.adobe.com
Vice Chair - UN/CEFACT Bureau Plenary - http://www.unece.org/cefact/
Adobe Enterprise Developer Resources  - http://www.adobe.com/  enterprise/developer/main.html
***********






-- 
***********
Senior Standards Strategist - Adobe Systems, Inc. - http:// www.adobe.com
Vice Chair - UN/CEFACT Bureau Plenary - http://www.unece.org/cefact/
Adobe Enterprise Developer Resources  - http://www.adobe.com/ enterprise/developer/main.html
***********






-- 

Internet Business Logic  --  online at www.reengineeringllc.com

Reengineering LLC,  PO Box 1412,  Bristol,  CT 06011-1412,  USA

Phone 860 583 9677     Mobile 860 830 2085     Fax 860 314 1029


<Prev in Thread] Current Thread [Next in Thread>