W3C home > Mailing lists > Public > www-ws-arch@w3.org > February 2003

Re: A Priori Information (Was Snapshot of Web Services Glossary )

From: Mahan Michael (NRC/Boston) <michael.mahan@nokia.com>
Date: Wed, 26 Feb 2003 17:08:25 -0500
To: "ext Cutler, Roger (RogerCutler)" <RogerCutler@chevrontexaco.com>, Assaf Arkin <arkin@intalio.com>, David Orchard <dorchard@bea.com>, Hugo Haas <hugo@w3.org>
CC: David Booth <dbooth@w3.org>, <www-ws-arch@w3.org>, Mark Baker <distobj@acm.org>
Message-ID: <BA82A509.37CE%michael.mahan@nokia.com>

Roger,

I believe 'a priori' and 'non a priori' are terms from the domain of agents
and semantic web. 'Non a priori' means 'without prior knowledge' and is used
to describe the property of an SW agent that can 'understand' what any
deployed and ontologically described service does/means and how to invoke
it, at runtime. In the context of AR023.7.1, this nuance is lost and so I
would recommend making the word change in 23.7.1 to 'prior' and not
bothering with the glossary.

Mike Mahan, Nokia


On 2/26/03 3:51 PM, "ext Cutler, Roger (RogerCutler)"
<RogerCutler@ChevronTexaco.com> wrote:

> 
> Geez, the WHOLE point of this is interpreting the phrase used in the
> charter and requirements in order to answer a formal issue.  The issue
> is what is meant when the phrase is used in the charter, and more or
> less copied into the requirements -- not what the phrase means in some
> general sense.  Is there any way we can found out who wrote the #$%%^#
> thing and ask what the heck they had in mind?
> 
> I remind you that even if we are thinking of the charter as governing
> law, the courts consider the intention of the legislative body when
> interpreting what a law means.  We don't exactly have to go back to the
> Federalist Papers to do this.
> 
> The change log of the Requirements Doc says something about "add Mark
> B's a priori requirement ...".  Does Mark Baker have something to do
> with the use of this phrase?  Mark Baker, in a posting 2/26, seems to
> say that he agrees that the term "prior" should replace "a priori".
> 
> Is there anyone who REALLY wants and cares about the use of "a priori"
> as opposed to "prior" in the charter and the requirements?  If not, can
> we possibly declare this argument to be moot?
> 
> -----Original Message-----
> From: Assaf Arkin [mailto:arkin@intalio.com]
> Sent: Wednesday, February 26, 2003 2:02 PM
> To: Cutler, Roger (RogerCutler); David Orchard; Hugo Haas
> Cc: David Booth; www-ws-arch@w3.org
> Subject: RE: A Priori Information (Was Snapshot of Web Services Glossary
> )
> 
> 
> 
> 
>> -----Original Message-----
>> From: www-ws-arch-request@w3.org [mailto:www-ws-arch-request@w3.org]On
>> Behalf Of Cutler, Roger (RogerCutler)
>> Sent: Wednesday, February 26, 2003 9:50 AM
>> To: David Orchard; Hugo Haas
>> Cc: David Booth; www-ws-arch@w3.org
>> Subject: RE: A Priori Information (Was Snapshot of Web Services
>> Glossary
>> )
>> 
>> 
>> 
>> I think that the term "a priori" really only has one rigorous meaning,
> 
>> and that is (as posted by Ugo 2/19):
>> 
>> A priori: relating to or derived by reasoning from self-evident
>> propositions.
> 
> That's one definition, but all the ones I've read [1] seem to imply that
> a priori knowledge is one that is not based on facts, such as
> self-evident, intuitive or whatever reasoning you had and want to
> validate.
> 
> So a use case would go like that:
> 
> - I think this service will give me the time of day (a priori)
> - I send a request, get a response, validate/invalidate my assumption
> 
> So, to further confuse the reader, if I had a definition of the service
> that I want to interact with, I would say that's a priori knowledge.
> It's not self evident or intuitive, but it's reasoning based on the
> service definition. It's not validated.
> 
> I send a request, get a response, validate my assumption based on that
> response and now my knowledge is no longer a priori.
> 
> On the other hand, if I have an end-point with no idea what it does,
> send a message and get some response back, then I learn what the service
> does without a priori knowledge. (End-point is considered prior
> knowledge in this
> context)
> 
> In other words, using a WSDL definition to interact with a service is a
> priori knowledge, but just sending some empty message and getting a
> response is no a priori knowledge with prior knowledge.
> 
> And before REST advocates interject, the same holds true for protocol
> listed in the URL. Assuming that a URL starting with http: implies an
> HTTP server at that end is a priori knowledge. We validate it by sending
> an HTTP request and seeing whether we get any HTTP response or some
> other response (no HTTP server here, go away!) or no response (no TCP
> server here, have a nice day).
> 
> arkin
> 
> [1] http://dictionary.reference.com/search?q=a+priori&r=3
> 
>> 
>> However, as David has pointed out, the objective here is really to
>> deal with the issue that has been raised.  We are, I believe,
>> suggesting that we don't really have to define the term in the
>> Glossary if we don't intend to use it in the architecture document,
>> but we do have to deal with the issue.
>> 
>> I think that the underlying assumption here, at least in my mind and I
> 
>> think in a number of others, is that the term was used somewhat
>> carelessly in the charter and that whoever wrote it really meant
>> "prior".  The statement is much easier to understand if one
>> substitutes "prior" for "a priori", and it seems to have a real
>> function that way. That is, if we interpret it in this way we think
>> that we are being consistent with, and responding to, the intended
>> meaning of the charter. I think that if we get embroiled in a lengthy
>> discussion of the term "a priori" we will, in fact, not be responding
>> to the intent of the charter.
>> 
>> At the very least, if we answer the issue in this way it puts the ball
> 
>> in the other court.  That is, if the framers of the charter REALLY
>> meant something other than "prior" they can tell us so, since we have
>> made it very clear that this is our best understanding of what they
>> meant at this moment.
>> 
>> If we take this path, however, I think we definitely do NOT want to
>> put a definition of "a priori" into the glossary that says it is
>> equivalent to "prior".  That would simply be propagating confusion.
>> 
>> -----Original Message-----
>> From: David Orchard [mailto:dorchard@bea.com]
>> Sent: Wednesday, February 26, 2003 10:47 AM
>> To: Cutler, Roger (RogerCutler); 'Hugo Haas'
>> Cc: 'David Booth'; www-ws-arch@w3.org
>> Subject: RE: A Priori Information (Was Snapshot of Web Services
>> Glossary
>> )
>> 
>> 
>> I dunno.  I think that the term "A priori" should be defined in a
>> rigorous way.  Can somebody summarize the differences between the
>> definitions that have been championed?
>> 
>> Dave
>> 
>>> -----Original Message-----
>>> From: www-ws-arch-request@w3.org
>>> [mailto:www-ws-arch-request@w3.org]On
>>> Behalf Of Cutler, Roger (RogerCutler)
>>> Sent: Wednesday, February 26, 2003 7:37 AM
>>> To: Hugo Haas
>>> Cc: David Booth; www-ws-arch@w3.org
>>> Subject: RE: A Priori Information (Was Snapshot of Web
>>> Services Glossary
>>> )
>>> 
>>> 
>>> 
>>> Well, the suggestion was NOT to put anything in the glossary for
>>> this term and to use the verbiage below as a response to the issue.
>>> 
>>> I'm not sure if we have anything explicit in the requirements about
>>> supporting late binding, but it seems to me that a number of people
>>> on
>> 
>>> the WG consider this important and that this was the sense of the
>>> statement in the charter.
>>> 
>>> -----Original Message-----
>>> From: Hugo Haas [mailto:hugo@w3.org]
>>> Sent: Wednesday, February 26, 2003 9:34 AM
>>> To: Cutler, Roger (RogerCutler)
>>> Cc: David Booth; www-ws-arch@w3.org
>>> Subject: Re: A Priori Information (Was Snapshot of Web Services
>>> Glossary
>>> )
>>> 
>>> 
>>> * Cutler, Roger (RogerCutler) <RogerCutler@chevrontexaco.com>
>>> [2003-02-24 10:41-0600]
>>>> OK, we've kicked this term around enough so that it seems
>>> pretty clear
>>> 
>>>> that it is not going to be a quick kill to get consensus on
>>> a general
>>>> definition, and I think David is absolutely correct: we need to
>>>> address the issue itself, but not necessarily this term as
>>> a general
>>>> concept.
>>>> 
>>>> So I suggest something along the following resolution to resolve
>>>> the
>>>> issue:
>>>> 
>>>> "The WG is not currently using the term "a priori
>>> information" in the
>>>> reference architecture, so we do not feel a need to come to an
>>>> agreement about the meaning of the term in general.  In the
>>> specific
>>>> context in which it is used in the group charter, we
>>> understand it to
>>>> mean "prior information".  We interpret this as a
>>> requirement that the
>>> 
>>>> architecture support late binding."
>>> 
>>> I am happy to put such a statement in the glossary. However, I think
> 
>>> that we should add something (or a placeholder) in the WSA to talk
>>> about it. Maybe just to say what you are saying here.
>>> 
>>> However, I was wondering if we had actually a requirement about this
> 
>>> before saying "We interpret this as a requirement that the
>>> architecture support late binding."
>>> 
>>> AC004 and AR004.2 read[1]:
>>> 
>>> |   AC004
>>> |          does not preclude any programming model.
>>> |
>>> |          + AR004.2 is comprised of loosely-coupled components and
>>> their
>>> |            interrelationships.
>>> 
>>> I think that this is the one that has been discussed when there were
> 
>>> late binding discussions, but I don't think that it explicitely
>>> calls out for it. Maybe we are missing a requirement then.
>>> 
>>> Or have I missed something in the requirements document?
>>> 
>>> Regards,
>>> 
>>> Hugo
>>> 
>>>   1. http://www.w3.org/TR/2002/WD-wsa-reqs-20021114#AC004
>>> --
>>> Hugo Haas - W3C
>>> mailto:hugo@w3.org - http://www.w3.org/People/Hugo/
>>> 
>>> 
>>> 
> 
> 
> 
> 
Received on Wednesday, 26 February 2003 17:07:47 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:15 GMT