OMG Internet Platform Special Interest Group
Minutes of Meeting #14
December 1, 1997
East Brunswick, New Jersey
OMG document internet/97-12-01
OMG Internet Platform SIG homepage: http://www.objs.com/isig/home.htm
Towards a Web Object Model, Frank Manola, Object
Services and Consulting
Experience with iM, a Componentized Web Infrastructure,
Pete Bonham, Director R&D/CTO Tektonic Software
Internet, Intranet, Extranet, Web: How should I Deploy
my CORBA Application, Imran Sayeed, CTO, NetNumina Solutions
HTTP-NG Briefing, Larry Smith, IBM
Brainstorm on OMG + Java Relationship,
led by Craig Thompson, Object Services and Consulting
Working Session on Object Transfer and Manipulation
(OTAM), led by Shel Sutton, The MITRE Corporation
CVW, Shel Sutton, The MITRE Corporation
Next Meeting Agenda
Dr. Craig Thompson - email@example.com - Object Services and Consulting,
Mr. Sheldon C. Sutton - firstname.lastname@example.org - The MITRE Corporation
Mr. Christophe Gransart - email@example.com - Laboratoire d'Informatique
Fond. de Lille
Dr. Philippe Merle - firstname.lastname@example.org - Laboratoire d'Informatique
Fond. de Lille
Mr. Jeff Poste - email@example.com - IPL Energy Works
Mr. Andrew Eisenberg - firstname.lastname@example.org - Sybase,
Mr. Aart van Halteren - email@example.com - KPN
Mr. Stig Berild - firstname.lastname@example.org - SISU
Ms. Susan Wolfley - email@example.com - The MITRE Corporation
Mr. David S. Dobrotka - USAF/AFIWC
Mr. Terry Ellis - Fretwell Downing
Ms. Catherine Susch - firstname.lastname@example.org - Ernst & Young
Mr. Benedicto Cabanas - email@example.com - The OASys Group
Ms. Kathleen Cowan - firstname.lastname@example.org - Freddie
Ms. Katharine Whitehead - email@example.com - SEER Technologies
Mr. Henry Balen - firstname.lastname@example.org - Xenotrope, Inc.
Mr. Pranab K. Baruah - email@example.com - Boeing Company
Mr. Denis L. Bagsby - firstname.lastname@example.org - Southwestern Bell
Mr. Antony Reynolds - email@example.com - Galileo International
Mr. Peter Bonham - firstname.lastname@example.org - Tandem Computers,
Mr. William Ruh - email@example.com - Concept Five Technologies,
Mr. Ron Zahavi - firstname.lastname@example.org - Concept Five Technologies,
Mr. John Marsh - email@example.com - Concept Five Technologies
Dr. Jeremy Parsons - firstname.lastname@example.org - European Bioinformatics
Mr. Imran Sayeed - email@example.com - NetNumina Solutions
Mr. Yoichi Mori - Yoichi.Mori@unisys.co.jp - Nihon Unisys, Ltd.
Mr. Silas Larry Smith - firstname.lastname@example.org - IBM Corporation
Mr. James L. Ellis - Merrill Lynch
Mr. Masayoshi Shimamura - email@example.com - Fujitsu
Mr. Jan R. Schultz - firstname.lastname@example.org - IDX Systems Corporation
Dr. Jeff Sutherland - email@example.com - IDX Systems Corporation
Mr. Edwin Tse - firstname.lastname@example.org - Ericsson
Dr. Umesh Bellur - email@example.com - Oracle Corporation
Mr. Simon C. Nash - firstname.lastname@example.org - IBM UK Ltd.
Mr. Ofer Ben-Shachar - email@example.com - NetDynamics
Mr. Peter Yared - NetDynamics
Mr. Vasile Radoaca - firstname.lastname@example.org - Nortel Technologies
Mr. Henry Rothkopf - email@example.com - The MITRE Corporation
Mr. Wang-Chien Lee - firstname.lastname@example.org - GTE Laboratories Inc.
Mr. Brian Thomas - email@example.com - Southwestern Bell
Mr. Lars Ola Osterlund - firstname.lastname@example.org
ABB Network Partner
Ms. Elizabeth Ungar - email@example.com - The Boeing
Ms. Maryann Spillane - firstname.lastname@example.org -
Raytheon Electronic Systems
Mr. Ralph Stout - email@example.com - Information Builders,
Mr. Rangaswamy Keshavan - Quoin
Mr. Michael Bigrigg - firstname.lastname@example.org - Carnegie Mellon University
Abstract: There are a number of Web technologies that are
relevant to providing object technology (and object services) on the Web
*in addition to Java*(!) This presentation reviews efforts to add more
structure to the Web such as XML, Document Object Model, PICS, and Resource
Description Framework. The presentation also discusses the key issue of
converging these technologies, and object efforts like IDL and Java, to
provide better structuring primitives for the Web, and avoid a situation
where each ommunity invents its own representations and services.
Thesis: There is a need to further integrate web and object technology.
The complexity of applications that the web is being applied to is increasing
and there is beginning to be overlap with OMG. OMA provides a useful architecture.
The notion that the web is a simple distributed object system is not new:
URLs provide OIDs, pages as state, behavior provided by HTTP servers and
scripting languages on client and server side. A number of efforts have
tried to amplify this integration. Object technology is a convenient language
for extending the web.
Fundamental components of any object model requires data structures
that can represent state and a means of associating behavior. Mapping this
to the web, we need to gather the data/pages and associated relevant code.
Java provides one means for doing all this. Two approaches: do it all in
Java or beef up parts of the web infrastructure.
W3C is doing relevant work.
How to form web objects using the technologies above: one reasonable approach
-- use XML as the state. Use Java tag for object code. Use RDF to define
relations to other code.
XML is an SGML subset. It is increasingly built into IE. Lots of end user
interest. Gave examples. Tags have application semantics, not necessarily
presentation semantics, which is separately specified. XML-Link, XML Stylesheets,
XML Namespace Facility (keep separate or mix the namespaces). See http://www.w3.org/XML/
PICS - Platform for Internet Content Selection - for defining content rating
by various third party rating services. Implemented as an http enhancement.
Ratings do not have to be embedded in the document. PICS-NG is like PICS
but uses XML so you can define your own tags. Some other XML data models
are XML-Data (Microsoft), Meta Content Format (Netscape), and Channel Definition
RDF - Resource Description Format - a standard way to represent metadata.
Many ways to associate assertions and resources. Can reify properties so
they can be assigned properties. RDF has defined serialization, assertions,
DOM - Document Object Model - provides an OO API, don't have to parse text.
Client can manipulate objects from client. DOM plug-ins. Class hierarchy
rooted at document with HTML and XML subtypes. You get a generic interface.
A purchase order is a generic object.
HTML supports OBJECT tag to generalize applets.
Final observations -- you can support many kinds of object models using
these mechanisms. All the representations have a predicate logic formulation.
Reification is important. Object logics provide a way to define query languages
over these structures.
Abstract: iM ( information Matrix ) is an application platform
- i.e. an architecture and one instantiation - for heterogeneous information
retrieval, transformation, and delivery. The iM architecture is based upon
coarse-grained CORBA components. iM has engendered a fast growing autocatalytic
product community. This presentation introduces iM, major architectural,
design, and implementation decisions, and then reviews lessons learned
to date and potential future work directions.
Notes on presentation: Tektonic is a business unit of Tandem.
Marchitecture: Tektonic does high performance IPC (fastest or close to
it). Krypton ORB on top. Distilling products off of iM. Started down this
path 4 years ago. Architecture must be extensible and scaleable - adapter,
gateway, transformers. (Note: start with clean architecture and then warts
appear which are secondary architectural abstractions). Few adapters, many
gateways. Services are protocol independent, stateless, not all services
you need are in your architectural space so gateways are critical. They
quickly made some design decisions that became irrevocable. No co-location
exploitation. Scaleable and replicable components. Parse headers into contexts
of attribute value pairs. All components in their architecture have IDL
interfaces so they can run in different environments. You need different
TCP components because they must be retargetable. Showed the stream interface
using get/put. Partners can extend the architecture via replacing modules
and server side scripting.
Lessons learned: coarse-grained asynchronous components yields an architecture
and implement. Might have reached a limit in componentry. The humble stream
still as legs. The component approach is best. Context is hard - seems
to be an optimization problem. Not so sure you can get top performance
w/o taking performance into account. CM, management, alerts is a big resource
sink. Need canonical approach to configuratiion and alerts. Having more
success via iscript as opposed to IDL.
Guesses and beliefs. Product will die if it does not form a solid community.
Not clear whose component problem it is in multi-component vendor environments.
Best component may be too heavy weight for someone else. Request/response
might need a dialog. Throughout the lifecycle, enormous productivity gains
are still possible.
Abstract: The industry is full of middleware, RAD tools, and
application servers that claim to be CORBA-compliant and able to work on
the Internet, Extranet, Intranet or the Web. Often the vendors have
no idea what the distinction is between these deployment scenarios and
why it's important. This presentation will detail the differring
technology requirements of these deployment scenarios, and offer solutions
using existing CORBA products, for some of their most vexing problems including:
The presentation will further explain an architecture for developing CORBA
applications that are well integrated with any of the above deployment
scenarios and conclude with a demonstration of a real CORBA application
developed using this architecture.
Running fully functional CORBA components in Web browsers
Efficient downloading of CORBA components
Notes on presentation: NetNumina moves business critical
applications to IDL and the web. What does web-enabling of an application
XML, … Use the web because it is omnipresent. CORBA is the only standards-based
distributed object infrastructure available in 100% pure Java. Much faster
to download Java ORB clients on Netscape and saves money on runtime licenses
since its built in. Can be bundled into jar or cab files for quicker downloads.
CORBA vendors are bundling new capabilities as jar files.
Internet-enabling is having apps available across firewalls.
Extranet traverses some walls Requires firewall, WANs, security
(authentication, authorization, data encryption, non-repudiation, integrating
w existing enterprise security). Intranet is inside the company-clients
want superior user interface and GUI controls.
CORBA provides support for Java. COM/CORBA bridge. CORBA services like
event, persistence, trader.
NetWeaver is a CORBA-based product that has CORBA clients that talk
SSL into web server that talks to CORBA, DCOM, DCE.
HTTP-NG Briefing, Larry
Larry Smith gave a 15 minute briefing on the W3C HTTP-NG project.
The World Wide Web Consortium (W3C, w3.org) initiates projects via briefing
packages. HTTP-NG is one such package. Two subprojects: web characterization
group and protocol design group.
The Web Characterization Working Group (chaired by Jim Pitkow) is working
on requirements, caching, hot spots, behavior of a high performance web.
They are trying to get good metrics on the health of the web and traffic
The Protocol Design Working Group (chaired by Jim Gettys) is testing
the hypothesis "Can a generic distributed object system be used as the
foundation of the Web?" The effort is tied into IETF. Other key people
are Bill Janssen, Henrik Nielsen, and Mike Spreitzer. Will be related via
RFP and IIOP at the transport layer. Xerox ILU is being used for prototyping.
In separate W3C work, IDL is being used in the DOM model.
Abstract. Open discussion on what it might mean for OMG to
be "Java-friendly" to help convergence (at least preclude divergence) of
the two communities. Could this mean:
client side protocol stack: transport - compression - … - browser
server side protocol stack: transport - …
(a) replace IDL with Java - nobody is suggesting this, are they?
Discussion. With OMG and Java recently becoming PAS submitters,
there is a danger that the standards could diverge if OMG and Javasoft
do not begin to make a concerted effort to converge. OMG management and
Javasoft management do not currently have a strategic relationship. But
there is an effort on the part of OMG submitters to align specifications
like Name Service, Java Beans, Call by Value, Component Model, and Java
(b) allow specs with interfaces in Java in place of IDL to be submitted
(c) encourage standard mappings of standard OMG specs to Java using
the IDL-Java mapping
(d) have OMG involved in some way in the Sun Java PAS
(e) other ideas
We briefly discussed a few of the alternatives listed above in the abstract
as well as a systematic replacement of dusty OMG specs like Trader, Query
Service and others that have never been commercially implemented. The discussion
was intense and spirited.
Action Item. At the Salt Lake City meeting, we will have a half
day morning session on OMG + Java relationship. In between, Craig Thompson
will issue a call for position statements (which may lead to email discussions).
Object Transfer and Manipulation (OTAM), Shel
Sutton, The MITRE Corporation
Abstract. OTAM is a proposed information access common facility
based loosely on ISO FTAM . It consists of a virtual file system that provides
an interface to a collection of physical file stores and database records.
A Trader stores the file and database schemas plus potentially more (e.g.,
data conversions). This is an attempt to objectify file systems so CORBA
can operate on them. FTAM can be viewed as a file system adapter.
Discussion. Shel handed out a White
Paper on OTAM (internet/97-12-05) and went over it. Mike Bigrigg
(CMU, email@example.com) volunteered to be document editor. Some work
assignments were made or reconfirmed. Expect more progress at the next
CVW, Shel Sutton,
The MITRE Corporation
Shel Sutton announced that MITRE is offering to provide OMG their Collaborative
Virtual Workspace (CVW) for use in organizing OMG activities. CVW is a
MUD that provides a rooms metaphor where avatars do all your work (a joke,
don't cha know). MITRE retains control of the server which contains shareware;
it is being rewritten in Java and will be free, Shel thinks. It is not
implemented using CORBA at present-- this might be a place for standardizing
CORBA interfaces. MITRE is fanning it out internally.
Next Meeting Plan
See Preliminary Agenda
for Meeting #15.