- From: Phil Tetlow <philip.tetlow@uk.ibm.com>
- Date: Sat, 3 Dec 2005 10:03:20 +0000
- To: SWBPD <public-swbp-wg@w3.org>
- Message-ID: <OF9B409A0F.B70BF4AE-ON802570CC.00372A2B-802570CC.00373D09@uk.ibm.com>
All,
Please find below comments on the OO Promer form Grady Booch.
Best Regards,
Philip Tetlow
Senior Consultant (Certified Technical Architect)
IBM Business Consulting Services
Mail: IBM United Kingdom Limited, 1175 Century Way, Thorpe Park, Colton,
Leeds, LS15 8ZB
Current Assignment: DWP BPRP (Metadata)
Mobile: +44 (0)7740 923328
Email: philip.tetlow@uk.ibm.com
----- Forwarded by Phil Tetlow/UK/IBM on 03/12/2005 10:02 -----
Grady
Booch/Boulder/IBM
@IBMUS To
Phil Tetlow/UK/IBM@IBMGB
29/11/2005 21:02 cc
Holger Knublauch
<holger@SMI.Stanford.EDU>
Subject
Re: *IBM Confidential: W3C OO Note
- I think that you will really like
this(Document link: Phil Tetlow)
I'm catching up on my email...
A few random comments on the document.
In the abstact it notes that "the resuability of domain models is often
limited because they are commmonly application-specific." this doesnt' make
sense to me: a domain model is by its very definition domain-specific and
ergo largely limited to that domain. if you meant "platform-specific" then
i can make sense of your statement: insofar as a domain model uses elements
that are specific to an implementation (ie assuming a LAMP stack or .NET or
some other technology) then reuse is indeed impaired. at any rate, the
fundamental issue is really a very direct one (namely, defining
abstractions with a clear separation of concerns) and wrapping that problem
in the language as named in the abstract somewhat obscures this underlying
problem.
in the abstract, I also vigorously disagree with your third assumption
(namely, that traditional oo applications are so closed); i rarely find
real oo systems that are able to have such a complete enumeration of all
such use cases. indeed - and this points back to another fundamental best
practice - the best oo systems are somewhat loose (and this is not inherent
in the oo-ness of those systems).
there's also an underlying assumption given in the abstact - and perhaps
this suggests my bias to the current direction of the semantic web work -
that there is "one" domain model. in my experience, in loosely coupled
systems, there are often several domain models with impedence matching
among them. this is simple pragmatics (and is manifest is many supply chain
systemms) in that - unless you are reaching of the most common medocrity -
it is not practical to have such tightly and centrally defined abstractions
(unless they are very very general).
in the introduction, it's somewhat misleading to say that almost all
software systems are centered around a domain model. while it may be true
that within a given development culture there exists a shared vocabulary
that may or may not be manifest in a formal domain model, said domain model
does not always form the core of a system's architecture (rather, that
vocabulary is often only indictentally projected into the implementation).
true, this is not necessarily the best practice, but the point of my
statement is that, pragmatically, many software systems are not architected
around an explicit domain model.
btw, the reason for my statement is that i don't have an enterprise
system-centric view of the world...there's a lot more software out there
than just such genres of systems.
re figure 1 - and i respect the simplification therein due to the intended
audience of this piece - i always have a hard time with domain models
written with no context: this design only makes sense in the context of
knowing the stakeholders that view this domain model. the example also goes
on to note that we may then turn this domain model into a suitable data
structure in our implementation. this may be so for simple systems, but it
is rare that that is "a" data structure that manifests this design: rather,
one may find database tables, xml schemas, and various classes that
collectively may reflect this domain model (or aspects of it) but
individually do not define "the" data structure. this is especially true of
distributed systems, in which there may be many componetized manifestations
of this domain model.
the rest of the document dives into the details of rdf schema and
owl...i've nothing meaningful to add to this bit (but my previous comments
suggest that the lead up to rationalizing these elements of the semantic
web are not quite correctly nor in a balanced manner presented. i will add
that the first part of the document is quite different in tone from the
second: while the former emphasizes the issues of abstraction best
practices, the second second somewhat bypasses these fundamentals and just
dwells upon the mechanics of the semantic web facilities.
Grady Booch
IBM Fellow
Voice: (303) 986-2405
Mobile: (303) 898-7091
Fax: (303) 987-2141
Video: (303) 795-6587/6626
GPS: 39.620/-105.076
Notes: Grady Booch/Boulder/IBM
E-mail: gbooch@us.ibm.com
Phil
Tetlow/UK/IBM@IBM
GB To
Grady Booch/Boulder/IBM@IBMUS
10/12/05 06:23 AM cc
Holger Knublauch
<holger@SMI.Stanford.EDU>
Subject
*IBM Confidential: W3C OO Note - I
think that you will really like
this
Grady,
Please find attached the latest note from the Semantic Web Software
Engineering Task Force (IMHO a superb document by Holger Knublauch, and
material that I plan to add to the first book coming out of the BPRP
program of work - due soon). This is a primer specifically aimed at Object
Oriented developers which brings home a number of ontology-driven ideas and
provides some much needed detail in this area. If you have any comments or
would like to contribute to the document directly, please feel free.
[attachment "OO Note Ver 1.0.htm" deleted by Grady Booch/Boulder/IBM]
Best Regards,
Phil Tetlow
___________________________________________________________
Senior Consultant (Technical Architect)
(Embedded image moved to file: pic29080.gif)
Contact Details:
(Embedded image moved to file: pic14429.gif) : Mail: IBM United Kingdom
Limited, 1175 Century Way, Thorpe Park, Colton, Leeds, LS15 8ZB
(Embedded image moved to file: pic26802.gif) : Current Assignment: DWP
CCM1 (BPRP WATCH2)
(Embedded image moved to file: pic22280.gif) : Work:N/A
(Embedded image moved to file: pic31125.gif) : Mobile: +44 (0)7740 923328
(Embedded image moved to file: pic30667.gif) : Fax: On Request
(Embedded image moved to file: pic10810.gif) : Email:
philip.tetlow@uk.ibm.com
___________________________________________________________
Attachments
- image/gif attachment: pic29080.gif
- image/gif attachment: pic14429.gif
- image/gif attachment: pic26802.gif
- image/gif attachment: pic22280.gif
- image/gif attachment: pic31125.gif
- image/gif attachment: pic30667.gif
- image/gif attachment: pic10810.gif
Received on Saturday, 3 December 2005 10:03:28 UTC