RE: what is discovery - One concrete proposal

+1 to Heather.

My intent was to call any process that provided a binding
the needed data , a discovery process. Conversations, phone calls, 
email or WSDL and UDDI all qualify.

The definition is loose enough to be able to say, discovery is
*always* present in web services. Ideally, it uses a mechanism that
allows the technical description data to be bound at runtime.
However, the binding data may be conveyed between parties informally, 
for example through oral conversations or email, and bound to running 
code statically at compile time or entered as a command line parameter,
etc. Regardless of the mechanism, the process of finding the service 
discription data needed to complete the binding is the process of
discovery.

If Heather's phrasing is clearer, then I like it.

Dave

-----Original Message-----
From: Jean-Jacques Moreau [mailto:moreau@crf.canon.fr]
Sent: Friday, October 11, 2002 6:18 AM
To: Heather Kreger
Cc: www-ws-arch@w3.org
Subject: Re: what is discovery - One concrete proposal



Heather,

I agree with everything you've said. My question was really 
whether what you call "static binding" is supported by Dave's 
definition.

Axiom 2 ("early") and axiom 4 ("was already know") suggest it 
might, but unless I've missed anything obvious, I don't get a 
clear answer from Dave's definition (but I get a clear answer 
from your message!).

Jean-Jacques.

Heather Kreger wrote:
> Discovery can be done before the client is developed, while the client is
> being developed (hence it is pre-compiled in). This is the most common
> scenerio being deployed today.  In IBM we call this static binding.
> 
> Discovery can also be done by the client at runtime. There are two types
of
> runtime binding:
> 1. The client has already discovered the interface, has programmed to it,
> and just the service instance (location from the WSDL) is discovered at
> runtime. This scenario is being deployed today, but less often than static
> binding.  WSIF supports this type of binding.
> 
> 2. The client discovers the interface specifics and the service instance
> during runtime. In the deployments of this that I know of, they use a DII
> style interface, like the JAXRPC call object or the WSIF apis to figure
out
> what message to create, create it and process the results.  There are not
> many of these out there.
> 
> Heather Kreger
> Web Services Lead Architect
> STSM, SWG Emerging Technology
> kreger@us.ibm.com
> 919-543-3211 (t/l 441)  cell:919-496-9572
> 
> 
> "Jean-Jacques Moreau" <moreau@crf.canon.fr>@w3.org on 10/11/2002 02:42:42
> AM
> 
> Sent by:    www-ws-arch-request@w3.org
> 
> 
> To:    Dave Hollander <dmh@contivo.com>
> cc:    www-ws-arch@w3.org
> Subject:    Re: what is discovery - One concrete proposal
> 
> 
> 
> 
> A SOAP client may use a service where the service information is
> obtained out of band (e.g. it may be precompiled into the
> client). Is this supported by your current definition or are you
> implying that discovery is mandatory (I don't read anything like
> discovery is optional)?
> 
> Jean-Jacques.
> 
> Dave Hollander wrote:
> 
>>To try to get temporary closure on the discovery,triangle,
>>and cloud, let me try to state one position.
>>
>>Recommendation:
>>1. Leave it in the spec dract as is or ammended with axioms
>>   from below.
>>
>>2. Add an example where "discovery" is a trivial role because
>>   there are two parties directly exchanging information that
>>   is hardwired into the service.
>>
>>3. Label the node "Discovery Agencies"
>>
>>--------------------------------------------------------------
>>
>>Discovery = exchange of the service description details necessary
>>to make a conncection.
>>
>>Discovery Axioms:
>>
>>1) discovery need not rely upon formal documents.
>>
>>2) discovery occurs regardless of when the discovered
>>    information is bound into the connection (early or late).
>>
>>3) discovery is discovery regardless if the provider or
>>    requestor does the advertising.
>>
>>4) discovery is discovery even if the data discovered was
>>    already known. All that needs to be true is the potential
>>    that the data *may* be different or new.
>>
>>5) discovery is discovery even if there are only two parties,
>>    requestor and provider.
>>
>>
>>I believe that "discovery", as defined above, exists as a
>>role in all of the scenarios that have been presented here.
>>
>>So that leads to the question: is "discovery", as defined above,
>>relevent enough to be included in our base architecture?
>>
>>I believe discovery is relevent and should be in the
>>base architecture for the following reasons:
>>
>>1. the distinction between hypertext and web services
>>   web has hypertext links to create a network, web
>>   services currently do not have a mechanism for defining
>>   a newtwork.
>>
>>2. good for the "ilities" (scalability, reliability, etc)
>>
>>3. it always happens, just sometimes it is done outside
>>   of the system.
>>
>>4. Most people expect to see it. If it is not there, our
>>   audience will either be disappointed or will try to find
>>   it. Either way confusion and mixed understanding will result.
>>
>>
>>
>>
> 
> 
> 
> 
>

Received on Friday, 11 October 2002 11:48:06 UTC