[Top] [All Lists]

Re: [ontolog-forum] Webby objects

To: "John F. Sowa" <sowa@xxxxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Fri, 16 Nov 2012 14:20:50 -0800
Message-id: <93632842-1303-4843-A23C-D2B6DD1897E5@xxxxxxx>
Just to correct the historical record:    (01)

On Nov 16, 2012, at 10:34 AM, John F Sowa wrote:    (02)

> On 11/15/2012 5:47 PM, Pat Hayes wrote:
>> I am as a matter of policy ignoring all rants about how
>> almost everyone got almost everything wrong...
> Pat, you are listed in one of the 22 groups that participated in DAML.    (03)

I was not aware that this large a number of groups "participated" in DAML. The 
actual DAML language, finally published as DAML+OIL after liaison with the 
European parallel OIL initiative, and almost immediately morphed into the W3C 
OWL initiative, was produced by a committee of approximately 25 people, of 
which I was indeed a member. (see http://www.daml.org/committee/ ). This 
committee, chaired by Mike Dean, acted very much like a W3C working group, with 
weekly teleconferences, active email discussions, ocassional F2F meetings, 
etc..     (04)

> My opinions are based on the original proposal, some interim reports,
> the final report, and various later documents.  I defer to you on
> any questions about what happened in that project.
> But that list also includes people as diverse as Roger Schank and
> Doug Lenat.    (05)

Neither Roger nor Doug had any significant input to the workings or discussions 
of the DAML committee. (It is news to me that Roger was even aware of it.) 
Cycorp was welcome to join the W3C consortium and send a representative to the 
WG. I believe in fact that they may have done so in later WGs. Like many 
people, they did not take DAML seriously at the time.    (06)

>  Tim's proposal [1] is sufficiently broad that the people
> on the list [2] could support it.  But I can't believe that all or
> even a majority of them would concur with the final report [3] and
> the subsequent developments.  Silence is not consent.    (07)

When it remains silence in the face of public calls for comments, a long 
process of open "last call" appeals for input and so on, all visible on a 
widely read and completely open public email discussion forum, then I believe 
that it can be treated as consent; indeed, that it must, in order for progress 
to be made.  As the marriage service says: let him speak now, or forever hold 
his peace.     (08)

> All my rants about the Semantic Web stem from two basic points:
>  1. Tim Berners-Lee's DAML proposal of February 2000 was mostly right
>     about the major issues required for the SW, but it was much more
>     ambitious than what could be implemented in five years.  But as
>     an outline and goal for the SW, it remains far better than
>     anything else that has been written about it.
>  2. The DL aficionados who promoted and enforced decidability pushed
>     the SW in a direction that was antithetical to nearly every major
>     issue that Tim had hoped to develop.  The slippery slope began
>     with OWL (full) from the DAML project,    (09)

The design of both DAML and OWL was indeed heavily influenced by DL enthusiasts 
(you can count them on http://www.daml.org/committee/ :-). And I agree with you 
that the emphasis on decideability has done more harm than good, although I now 
think it less important than I once did. But OWL Full is not DL based, is not 
decideable, and is not a leftover from DAML. I feel rather strongly about this 
because OWL Full is essentially my invention, produced in the teeth of often 
heated opposition from the DL enthusiasts on the WG, who had written detailed 
memos proving that it was impossible (because of the Russell paradox). OWL-DL 
is, however, far and away the most widely used version of OWL precisely because 
it comes with effective reasoners with guarantees of performance.     (010)

> but it went over the cliff
>     with the later emphasis on decidability.    (011)

This was not "later": it has been a core idea of the whole DL movement from the 
start, and has always been an important theoretical issue in the DAML and OWL 
discussions. Ian Horrocks is of the view that only a decideable logic can 
really be called a logic, and that classical logics are simply inadequate to 
capture the richness of human reasoning. I (of course) disagree, but one cannot 
simply dismiss Ian as a lunatic: his viewpoint does have some traction in the 
real world of data. I have often found it very hard to explain to people what 
it means to be complete yet undecideable, for example. This distinction does 
not sit lightly in most people's minds.    (012)

> Everything else supports or follows from those two points:
>  1. Three terms that Tim emphasized in the DAML proposal were
>     diversity, heterogeneity, and interoperability.  He discussed
>     those points and the requirements for supporting them.  He also
>     said that trust had to be addressed "early" in the DAML project.
>  2. In the 1990s, the most widely used DL systems were LOOM and
>     PowerLOOM.  They were hybrids with a DL for the type hierarchy
>     and with lots of features that users needed for application
>     development.  Bob MacGrgegor (who was also on the DAML list)    (013)

Bob McGregor gave no input to the DAML or OWL process whatsoever, but 
complained loudly after the proposals were published as recommendations. He is 
my prototypical example of why silence has to be taken as consent.     (014)

>     stated the obvious:  "Users always ask for more expressive
>     power, and they never ask for decidability."    (015)

I wish this were true. Unfortunately, there are many users who either ask for, 
or worse simply *presume* decideability. So that if you give them a system 
which sometimes says "no answer", they will simply take this as the answer, 
"no". Which is one of Ian's arguments: whatever you give users, they will treat 
it as an oracle. So it had better *be* an oracle, to avoid trouble.     (016)

>  3. Tim also mentioned a wide range of logics and systems, which
>     included KIF, Prolog, and relational databases.  From [1], "the
>     Semantic Web must provide a common interchange language bridging
>     these diverse systems...  SWeLL (Semantic Web Logic Language)
>     will enable this interchange."  SWeLL is the "unifying language
>     for classical logic", and the layer cake shows that SWeLL
>     includes propositional, first order, and higher order logic.
>  4. In the final DAML report, SWeLL became N3, which is a simpler
>     notation for RDF triples    (017)

RDF does not impose any particular notation for triples. Right now, Turtle is 
probably more standard than RDF/XML. The diagrams produced by COE are a 
perfectly valid and conformant notation for RDF and OWL.     (018)

> , but it also goes beyond RDF in
>     expressive power (exactly what power that includes is not
>     stated).  But Figure 2, which shows "The Semantic Web Wave",
>     puts N3 under the heading "non-web or non-standard".  Even
>     worse, SQL, KIF, Prolog, Cyc, and "etc" are in the pre-1998
>     "wave" -- legacy systems that don't even reach the level of
>     "non-web or non-standard".    (019)

All this talk of "waves" is just window dressing for the DARPA managers and 
others of limited intellectual power. At that time, there was a strong sense in 
the SW community that there was a need to "sell" the SWeb idea to people, and 
this kind of BS was one way to do that. But nobody should take such stuff 
seriously, particularly a decade after the event.     (020)

>  5. SQL runs the world economy, it's an ISO standard, and it is still
>     the primary DB language for commercial web sites.  To claim that
>     SQL is "non-web or non-standard" pushes the overwhelming majority
>     of commercial web sites outside the scope of the SW.  That is why
>     Google, Microsoft, Apple, IBM, and Amazon do not use OWL.  They
>     might import RDF, but their tools immediately convert RDF to more
>     efficient notations.    (021)

That is fine. RDF is designed and intended to be a language for interoperation 
and communication, not for processing. If anyone has their own in-house system 
or notation they prefer to use, and can translate into and out of it, that is 
precisely what the RDF authors intended to happen. It is not a refutation of 
the RDF vision.    (022)

>  For Google, JSON is the notation used by
>     all Google apps.  IBM's Watson uses SQL for its DB, it uses Prolog
>     for complex AI reasoning, and it uses XML for the UIMA notation,
>     not for RDF or OWL.    (023)

David Ferrucci told me that it uses a wide variety of systems and notaiton, 
including RDF and OWL. Admittedly not widely, but it does use them.     (024)

>  6. In the yearly interim reports for DAML, there were fewer and
>     fewer references to Tim's points about diversity, heterogeneity,
>     interoperability, and trust.  Diversity and heterogeneity are
>     not mentioned in the final DAML report.    (025)

Since DAML's job was to invent a single notation, what would you have had it 
say about diversity and heterogeneity? All these groups are given a charter and 
have to stay within it. Right now, the RDF2 WG has a very narrow charter which 
explicitly prohibits certain kinds of change form being made to the language. 
Constraints like this, and the need to continually compromise, are simply the 
facts of life in writing standards documents and specifications. One does what 
one can.      (026)

>  Interoperability occurs
>     only in the sentence "N3 is used as an interlingua for interoper-
>     ability for different systems", but Figure 2 says that N3 is
>     "non-web or non-standard".  And trust "continued to exist only
>     at the research level."
>  7. The requirement of decidability eliminated heterogeneity and
>     diversity.  N3, RuleML, SWIRL, and SWeLL were eliminated.
>     PowerLOOM, Cyc, Prolog, and every major programming language
>     are undecidable, and their users would never accept anything
>     less.  DLs have always been most useful as part of a hybrid
>     with more expressive languages.  But the decidability purists
>     forced every one of them out of the approved list.    (027)

Most of the legacy systems were classified as "pre-web" by virtue of their not 
using URIs as identifiers. And the DL enthusiasts on the DAML and OWL groups 
were the ones who were rooting for more conformity with legacy systems, and 
against the use of RDF (and especially RDF/XML). It was not decideability that 
eliminated talk of "heterogeneity and diversity" but rather a sense that these 
words didn't really amount to anything definite and "sellable". (If the SWeb is 
just the sum total of existing systems, then we have that already, so what is 
the fuss about?) And in fact, now, over a decade later, diversity and 
heterogeneity are a problem rather than a design principle. The fact is, most 
of the useful data out there is starting to get rendered into RDF, and when 
that is done, it is immediately useful and almost immediately made use of. And 
until that is done, it is not of the slightest use to me and I am sure to a 
large number of other people.     (028)

Pat    (029)

IHMC                                     (850)434 8903 or (650)494 3973   
40 South Alcaniz St.           (850)202 4416   office
Pensacola                            (850)202 4440   fax
FL 32502                              (850)291 0667   mobile
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (030)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (031)

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