[infod-wg] Notes from INFOD F2F on Saturday 25 June 2005
Susan Malaika
malaika at us.ibm.com
Mon Jul 4 23:26:15 CDT 2005
(Notes from remainder of F2F will be posted soon)
Attendees
- Stephen Davey (on Monday), Vijay Dialani (phone Saturday and Sunday),
Abdeslem Djaoui, Steve Fisher, Dieter Gawlick, Chris Kantarjiev, Cecile
Madsen, Susan Malaika, Mallikarjun Shankar
Agenda Saturday:
- create new scenario for sensors
- review R-GMA scenario - diagrams, operations
- close some existing open issues (consumption description and role of
subscription in particular)
Proposed Agenda for Sunday and Monday:
- agree notes from Saturday
- agree way forward
- complete RGMA scenario
- summarise list of specification open issues
- review nextGrid Animation scenario updates
- review latest wsn, including the mapping with infod, and the pullpoint
discussion
- create review presentation materials for infod and for ogsa data
sessions (especially for ogsa data transfer and replication)
- consider replication as a scenario
Actions:
- Arjun:
- Place presentation in GridForge INFOD Current Drafts
- Work on Use Case in two parts - Current; Future
- Send working group the link to the trade paper on SensorNet
- Steve Fisher:
- Add the 13 Rules section to the Use Case Template
- Update RGMA use case to include these operations (or whatever is
decided)
1. createPublicationType('NetworkTCPThruput', 'sql92',
'create table thruput...') return an EPR
2. createPublisher('site1') - also returns an EPR
3. AEPR = associate(PEPM, PTEPR, 'RP=5mn', 'rgma',
'incpomlete...')
4. createSubscription(PTEPR, policy, 'r=gma',
'select ...')
Items agreed upon (done) so far:
[Which specification issues do these relate to? Are these new issues which
were agreed? old closed issues that were re-opened? or old issues that
were still open?]
1. Define createPublication instead of generic createINFODResource
operations
2. Introduce createAssociation new operation with parameters: publisher,
publicationType, policy
3. Introduce getAssocation new operation (query associations)
4. Introduce dropAssociation
5. Ensure that no updates of publicationType or publishers allowed
6. Ensure that publicationType is used to define schema vocabulary
(doesn't include policy anymore)
7. Ensure that no updates of association allowed
8. Subscriptions do not depend on an explicit existing publications (with
or witout policies) because the subscribers may subscribe before
publishers register themselves or before publishers declare the QoS terms
or policies they support
Issues to be added to specification issues list
Issue 1: Should a concept of a producer be used to represent a sensor
Issue 2: Should explicit aggregation requests be part of the spec? Should
transformations be part of the spec
Issue 3: Should INFOD have recommendations on granularity of publishers?
Issue 4: Does a publisher have to register itself
Issue 5: Does a publisher have to support subscriptions?
Issue 6: Review these relationships:
INFODPublisher <-> INFODProducer <-> INFODSource
INFODSubscriber <-> INFODConsumer <-> INFODSink
...........................................................................................................................................................................................................
Here is the informal chat that accompanied the session
[10:41:25 AM] malaikas says: Hello
[10:41:43 AM] *** malaikas added Cec Madsen......................, Chris
Kantarjiev to this chat ***
[10:42:10 AM] malaikas says: Hello
[10:42:45 AM] *** Chris Kantarjiev added Steve Fisher to this chat ***
[10:43:36 AM] malaikas says: Hello Steve
[10:43:41 AM] malaikas says: Hello Chris
[10:43:47 AM] malaikas says: Hello Abdeslem
[10:43:52 AM] malaikas says: Hello Cecile
[10:44:12 AM] *** Chris Kantarjiev added Vijay Dialani to this chat ***
[10:44:23 AM] malaikas says: Hello Vijay
[10:44:27 AM] *** Chris Kantarjiev added Dieter Gawlick to this chat ***
[10:44:29 AM] Vijay Dialani says: hi
[10:44:34 AM] malaikas says: Hello
[10:47:02 AM] Steve Fisher says: Hello
[10:47:45 AM] malaikas says: these are the names I'm giving to security
today so you can enter the building at the weekend:
- Stephen Davey
- Abdeslem Djaoui
- Steve Fisher
- Dieter Gawlick
- Chris Kantarjiev
- Cecile Madsen
- Mallikarjun Shankar
Our meetings will be held on the fourth floor. I'll leave information with
security on how to reach me and get entry to the meeting rooms. You can
also reach me by email or instant messaging
The goal of the meeting is to drive closure on the specification open
issues, and thus update specs accordingly. Examing the use cases and WSN
will form part of that process.
Agenda items will include (the order will be decided at the F2F):
- summarise list of specification open issues
- review R-GMA scenario - diagrams, operations
- review nextGrid Animation scenario updates
- create new scenario for sensors
- consider replication as a scenario
- review latest wsn, including the mapping with infod, and the pullpoint
discussion
- close some existing open issues (consumption description and role of
subscription in particular)
- review presentation materials for infod and for ogsa data sessions
(especially for ogsa data transfer and replication)
See you on Saturday or at GGF14 next week :)
[10:53:38 AM] malaikas says: Sat - Sensors,
[10:53:44 AM] malaikas says: Sun - WSN
[10:53:53 AM] malaikas says: Sat - RGMA
[10:54:09 AM] malaikas says: Sun NextGrid
[10:54:29 AM] malaikas says: Sat - Summarise spec issues - start comiling
list
[10:54:43 AM] malaikas says: Sun - Replication
[10:55:01 AM] malaikas says: Sun - Review Presentation Materials
[10:55:20 AM] malaikas says: Sun - Create and Review Presentation
materials
[10:55:52 AM] malaikas says: Sat - Sensors RGMA Spec Issues
[10:59:32 AM] malaikas says: Arjun : Sensornet Background
[11:00:13 AM] malaikas says: A WAN Infrastructure to support sensor data
management
[11:00:22 AM] malaikas says: An Internet for Sensors
[11:00:34 AM] malaikas says: The Internet of Things analogy (RFID)
[11:01:18 AM] malaikas says: Goals, Prototyping, Info Dissemination
scenarios
[11:01:31 AM] malaikas says: We: = Oakridge
[11:02:36 AM] malaikas says: Greater safety in cities (chemical ..... ,
emergency responders, situational awareness)
[11:03:00 AM] malaikas says: We includes congress
[11:03:58 AM] malaikas says: Oakridge, Los Alamos, Laurence Livermore,
Laurence Berkeley, Brookhaven, Argonne
[11:04:18 AM] malaikas says: Pacific National Lab .... Did I get them all?
[11:04:39 AM] malaikas says: Contract research for the government
[11:05:05 AM] malaikas says: Don't compete with Industry
[11:05:20 AM] malaikas says: Don't as theoretical as academics
[11:05:47 AM] malaikas says: We: Federal Government - Naval connections
for sensornet
[11:07:09 AM] malaikas says: Interdisciplinary - Billion dollars budget
[11:07:33 AM] malaikas says: organised in major directorates
[11:07:44 AM] malaikas says: Arjun in computational sciences
[11:08:24 AM] malaikas says: Oakridge - goal to have fastesst
supercomputer in the world
[11:09:22 AM] malaikas says: Battelle non profit in Ohio
[11:10:31 AM] malaikas says: Bell's - Number of computing devices growing
log scale
[11:10:55 AM] malaikas says: Batelle, Lockheed, Uni of California etc
manage the various labs
[11:11:42 AM] malaikas says: Place sensors around US
[11:11:46 AM] malaikas says: Connect them
[11:12:21 AM] malaikas says: Read data, Infer interesting things
[11:12:55 AM] malaikas says: Give advice
[11:14:11 AM] malaikas says: Use existing standards to communicate between
middleware and sensors
[11:14:25 AM] malaikas says: IEE4051
[11:14:49 AM] malaikas says: Dieter: Each definition is a vocabulary
[11:15:24 AM] malaikas says: IEEE 1451
[11:15:51 AM] malaikas says: Sensor ML is converging to 1451
[11:16:36 AM] malaikas says: OGC (Open Geospacial) -
[11:16:54 AM] malaikas says: http://ieee1451.nist.gov/
[11:17:27 AM] malaikas says: http://www.opengeospatial.org/
[11:18:11 AM] malaikas says: Between sensors and applications -> Web
Services
[11:18:47 AM] malaikas says: Comment from Arjun : Publisher ->
Disseminator ; Subscriber, Consumer
[11:23:57 AM] malaikas says: Issuue: Do we need a concept of a producer to
represent a sensor
[11:25:42 AM] malaikas says: Question: Does IEEE 1451 define the interface
between the sensor and the publisher?
[11:27:33 AM] malaikas says: Wired devices are aggregating material
published by sensors
[11:29:51 AM] malaikas says: How deep should a query go?
[11:30:50 AM] malaikas says: Oracle edge server
[11:31:25 AM] malaikas says: Issue: INFOD recommendations on granularity
of publisher
[11:32:40 AM] malaikas says: Arjun Issue: The spec discusses information
that is collected
[11:33:31 AM] malaikas says: Arjun Issue: Explicit aggregation request
should be outside the spec - also transformations outside spec
[11:36:03 AM] malaikas says: Issue 1: A concept of a producer to
represent a sensor
[11:36:50 AM] malaikas says: Issue 2: Explicit aggregation request -
should that be part of the spec? Should transformations be part of the
spec
[11:37:33 AM] malaikas says: Issue 3: Should INFOD have recommendations on
granularity of publishers?
[11:39:07 AM] malaikas says: RFID -> identifier
[11:39:23 AM] malaikas says: Sensor -> data gatherer
[11:39:47 AM] malaikas says: Sensornet -> middleware (OGC)
[11:40:39 AM] malaikas says: Sensornet is a publisher
[11:41:09 AM] malaikas says: Sensornet node is a publisher'
[11:42:09 AM] malaikas says: Disseminators are configured at the system
startup
[11:42:57 AM] malaikas says: Node supports 1451;
[11:43:13 AM] malaikas says: Disseminator supports OGC
[11:43:21 AM] malaikas says: Node is sensitivie to location
[11:44:24 AM] malaikas says: Web Server at the node
[11:46:14 AM] malaikas says: Data Store to collect info from sensors
[11:46:28 AM] malaikas says: 20gigs of disk for a store
[11:47:16 AM] malaikas says: Tomcat; edge server environment
[11:48:07 AM] malaikas says: Actuator is a consumer
[11:48:10 AM] malaikas says: ?
[11:49:58 AM] malaikas says: [actuator sensors ] -> transducers
[11:51:03 AM] malaikas says: In OGC : Have features
[11:51:17 AM] malaikas says: First responders ?
[11:53:04 AM] malaikas says: Patterns in archives - temporal patterns
[11:59:28 AM] malaikas says: Discussion about Trucks scans and preventing
crime
[12:01:29 PM] malaikas says: Fall Deployment plan - place nodes in
Washington
[12:02:58 PM] malaikas says: Access control of data ..... firedepartment,
police, sharing information
[12:03:22 PM] malaikas says: Interdiction - Preventative Action
[12:05:12 PM] malaikas says: Keyhole bought by google
[12:06:24 PM] malaikas says: Hooks in keyhole to overlay information
[12:10:21 PM] malaikas says: do u have the charts vijay?
[12:11:33 PM] malaikas says: Data sharing services
[12:11:56 PM] malaikas says: Common Alerting Protocol (Homeland Security
supports)
[12:12:29 PM] malaikas says: Publication message could conform to CAP in
some applications
[12:13:44 PM] malaikas says: Sensor data -> Active Archive: Traffic Data
(Ontology) -> Services -> Consumers
[12:15:28 PM] malaikas says: Get Data for consumption -> Pull
[12:15:37 PM] malaikas says: Get Data -> for metadata
[12:16:55 PM] malaikas says: GetDataforConsumption -> Disseminator
[12:17:01 PM] malaikas says: GetData -> Registry
[12:17:19 PM] malaikas says: Applications receive data through pull
[12:19:32 PM] malaikas says: Issues that SensorNet have
[12:20:19 PM] malaikas says: Naming, Representation, placement.caching...
[12:21:29 PM] malaikas says: Developer Use Cases ...
[12:22:40 PM] malaikas says: (1) Common Opernational Model -> View
Sensornet assets in a region -> Manipulate Overlay
[12:25:34 PM] malaikas says: (2) Get all sensor data
[12:27:12 PM] malaikas says: Data center for data and for registry info
[12:27:34 PM] malaikas says: Regional data centers and national data
centers
[12:28:15 PM] malaikas says: Features means both the data and metadata
[12:32:31 PM] malaikas says: Question: Where does data get stored when
computed?
[12:40:00 PM] malaikas says: Question: Republish? Can anyone publish?
[12:42:39 PM] malaikas says: Issue 1: A concept of a producer to
represent a sensor
Issue 2: Explicit aggregation request - should that be part of the spec?
Should transformations be part of the spec
Issue 3: Should INFOD have recommendations on granularity of publishers?
Issue 4: Does a publisher have to register itself Issue 5: Does a
publisher have to support subscriptions?
[12:46:39 PM] malaikas says: Discussion: Publisher <-> Consumer ; Producer
<-> Subscriber
[12:47:00 PM] malaikas says: subscriber is an agent of the consumer
[12:47:30 PM] malaikas says: Publisher <-> Subscriber
[12:47:42 PM] malaikas says: Producer <-> Consumer
[12:47:53 PM] malaikas says: Publishers and Subscribers are clever
[12:48:04 PM] malaikas says: Producers and Consumers can be dumb
[12:48:42 PM] malaikas says: Abdeslem : Changed the terminology
[12:49:36 PM] malaikas says: Source + Producer (intead of publisher) <->
Consumer + Subscriber
[12:59:52 PM] malaikas says: Issue 6: Review - INFODPublisher <->
INFODProducer <-> INFODSource + INFODSubscriber <-> INFODConsumer <->
INFODSink
[1:01:34 PM] malaikas says: Issue 1: A concept of a producer to
represent a sensor
Issue 2: Explicit aggregation request - should that be part of the spec?
Should transformations be part of the spec
Issue 3: Should INFOD have recommendations on granularity of publishers?
Issue 4: Does a publisher have to register itself Issue 5: Does a
publisher have to support subscriptions? Issue 6: Review - INFODPublisher
<-> INFODProducer <-> INFODSource + INFODSubscriber <-> INFODConsumer
<-> INFODSink
[1:08:55 PM] malaikas says: Action: Steve Fisher to add the 13 Rules
section to the Use Case Template
[1:12:52 PM] malaikas says: Discussion of subscribers chooses publications
[1:12:57 PM] malaikas says: Rule 0
[1:18:43 PM] malaikas says: Issue 7: How to model including multiple
publishers in a subscription ? (should it be an in-line view?where
multiple sources are listed)
[1:21:10 PM] malaikas says: Proxies for applications in data centers -
selections by name and data types and region
[1:22:03 PM] malaikas says: Write 5 subscriptions for the 5 sources
[1:31:55 PM] malaikas says: Action: Data Model for the registry -
[1:38:18 PM] malaikas says: Requirements - Arjun did not understand R4
[1:42:31 PM] malaikas says: Arjun: Spec is about mechanism
[1:46:17 PM] malaikas says: Issue 8: How does the registry determine the
notifications it should issue to inform the appropriate parties of
relevant changes?
[2:02:29 PM] malaikas says: Action: Arjun PLace Presentation in GridForge
INFOD Current Drafts, Work on Use Case (Current) (Future), Send workgroup
the link to the trade paper on SensorNet
[3:03:56 PM] malaikas says: Lunch:
[3:04:07 PM] malaikas says: Chris introduces model
[3:04:39 PM] malaikas says: Terminology: Publication -> Publication Type
[3:05:44 PM] malaikas says: Entity has processing rules ; Artifact
[3:06:08 PM] malaikas says: Everything is an artifict
[3:08:58 PM] malaikas says: Publisher is an entity
[3:09:30 PM] malaikas says: Publicationtype ia artifact
[3:09:47 PM] malaikas says: Publicationtype has a transormation type
[3:13:21 PM] malaikas says: Publicationtype has a transformation
[3:16:14 PM] malaikas says: Action: Cecile - add Arjun to author list
[3:17:43 PM] malaikas says: Steve introduces RGMA
[3:18:27 PM] malaikas says: 3 audience: Site Admin
[3:19:17 PM] malaikas says: (1) Side admin (2) Network analyst (backbone)
(3) Resource Broker
[3:19:52 PM] malaikas says: Resource Broker -> ensure appropriate QoS
[3:20:34 PM] malaikas says: Nodes in the network with probes (sensors) -
probes could be wrappered grid FTP
[3:21:07 PM] malaikas says: Resource Broker decides where app is run or
where data is moved to or both
[3:22:10 PM] malaikas says: Site has multiple storage a compute resources
- and has a site admin
[3:22:24 PM] malaikas says: Sites are publishers
[3:22:49 PM] malaikas says: CERN is a site; Ruther ford
[3:24:25 PM] malaikas says: Publisher= producer = machine
[3:26:36 PM] malaikas says: Registry:
[3:26:41 PM] malaikas says: Vocabulary relation
[3:26:57 PM] malaikas says: Vocabulary is relational
[3:28:18 PM] malaikas says: Relational Schema
[3:28:32 PM] malaikas says: Tablenames, types, columns
[3:29:37 PM] malaikas says: One publication type or 7 publication types
[3:29:49 PM] malaikas says: 7 publication types agreed
[3:36:24 PM] malaikas says: Publisher source -> INFOD Description
[3:39:41 PM] malaikas says: Core : infod type, infod identifier ,
infod_name, infod_resourcereference ; constraints: infod_description;
infod_data;infod_descriptionfilters;infoddatafilters
[3:41:27 PM] malaikas says: Scheduling: infod_scheduling; policies:
infod_policies; processing: infod-processing
[3:58:29 PM] malaikas says: Tasks : (1)
CreateInfodResource(publicationtype,...) - insert ignore
(2)CreateInfodResource(publisher,site1..) and constraints [list of 7
types]
[4:00:04 PM] malaikas says: Arjun proposal: Register publishers first
[4:00:38 PM] malaikas says: Conclusions: Enforce publisher creation before
publication
[4:01:14 PM] malaikas says: Strike conclusion ....
[4:07:15 PM] malaikas says: Tasks : (1)
CreateInfodResource(publicationtype,...) - insert ignore
(2)CreateInfodResource(publisher,site1..) and constraints [list of 7
types] (3) UpdateInfodResource (for joinmatching)
[4:10:04 PM] malaikas says: Conclusion: UpdateInfodResource ensure
publisher exists
[4:12:04 PM] Vijay Dialani says: i have a question
[4:12:10 PM] malaikas says: Tasks : (1) CreatePublicationType
(2)CreatePublisher(3) Associate
[4:12:11 PM] malaikas says: Ask
[4:12:20 PM] malaikas says: Ask the question
[4:12:23 PM] malaikas says: we can hear
[4:12:28 PM] malaikas says: or type the question
[4:12:35 PM] Vijay Dialani says: ok i will speak in a moment
[4:12:41 PM] malaikas says: type it here
[4:13:22 PM] malaikas says: we cant hear if u r speaking
[4:14:16 PM] Vijay Dialani says: If we have createpublicationtype...we are
allowing the publishers to either use the new publication or the publisher
may createPublisher and just associate with publishertype...So what
happens to infod descriptions...is it true that now descriptions are only
associated with publisher
[4:17:16 PM] malaikas says: Very good point vijay
[4:17:48 PM] malaikas says: so a publication description swill apply to
all publishers publishing that publication
[4:18:23 PM] malaikas says: an example : publication description= only
published between 1am and 2am .....
[4:18:46 PM] malaikas says: so the publishers are required to publish the
publication between 1am and 2am
[4:19:05 PM] malaikas says: It is not clear tho who enforces that
publication constraint
[4:19:26 PM] malaikas says: is it the publisher who enforces the
publication constraint?
[4:19:32 PM] malaikas says: what do u think Vijay?
[4:21:04 PM] malaikas says: make publicationtype immutable
[4:21:06 PM] malaikas says: ?
[4:22:50 PM] malaikas says: Publications are handled by the registry
[4:24:52 PM] malaikas says: Publication constraints are just intents
[4:26:03 PM] malaikas says: Vijay: conclusion - so registry takes control
the lifetime of the publication after it is created
[4:26:39 PM] malaikas says: and manages who can delete or update the
oublication
[4:26:49 PM] malaikas says: Vijay : How does the registry decide?
[4:27:13 PM] Vijay Dialani says: no..no one can modify the
publicationtype...it can only be deleted
[4:27:25 PM] malaikas says: ok it is immutable as you said
[4:27:36 PM] malaikas says: OK - I will make sure everyone really agrees
[4:28:28 PM] Vijay Dialani says: registry manager decides when to delete
[4:28:45 PM] malaikas says: Conclusion: types are immutable cannnot be
changes
[4:28:51 PM] malaikas says: changed
[4:30:57 PM] malaikas says: So Vijay - PublicationsTypes and Publishers
cannot be changed
[4:31:03 PM] malaikas says: Do you agreee?
[4:32:14 PM] Vijay Dialani says: true
[4:32:18 PM] malaikas says: good
[4:33:43 PM] malaikas says: Now the associations are being defined as
separate resources
[4:35:08 PM] malaikas says: associations are updateable - do you agree
vijay?
[4:35:19 PM] Vijay Dialani says: yes
[4:35:53 PM] malaikas says: Good
[4:35:55 PM] malaikas says: Shout
[4:48:06 PM] malaikas says: Conclusions: (1) CreatePublicationType,
Publisher (no more Create resourcetype) ; (2) CreateAssociation; (3)
RestrictUpdates to associations
[4:49:34 PM] malaikas says: (4) Create Subscription
[5:08:35 PM] Vijay Dialani says: is the current issue last agenda item for
the day?
[5:09:04 PM] malaikas says: doubt it
[5:09:22 PM] Vijay Dialani says: tea break!!
[5:10:18 PM] malaikas says: we dont take tea breaks
[5:31:30 PM] malaikas says: Vijay: - (1) Publisher states what policies
s/he likes - (2) consumer chooses from the policies in the registry
[5:31:36 PM] malaikas says: did i say it right?
[5:31:43 PM] Vijay Dialani says: yes
[5:31:43 PM] malaikas says: can u fix what i wrote vijay?
[5:31:56 PM] Vijay Dialani says: ok..here is my view
[5:37:49 PM] Vijay Dialani says: policy should be mapped into the registry
for efficient evaluation
[5:48:40 PM] malaikas says: Policy is immutable
[6:17:58 PM] Vijay Dialani says: away for a minute
[6:18:59 PM] malaikas says: ok
[6:19:24 PM] Vijay Dialani says: back here
[6:24:14 PM] malaikas says: ok
[6:25:06 PM] malaikas says: so we are talking about publication type (or
assocation) having policy and schema vocabularies
[6:25:50 PM] malaikas says: Pubicationtype -> {schemavocab, policyvocab}
[6:26:13 PM] malaikas says: vocab is a top level directory thing
[6:26:26 PM] malaikas says: i mean vocab is a top level registry thing
[6:27:05 PM] malaikas says: Publicationtype -> {schemavocab}
[6:28:13 PM] malaikas says: Assocation {Publicationtype, Publisher,
policyvocab}
[6:30:44 PM] malaikas says: Association ->{Publicationtype, Publisher,
policyvocab}
[6:36:52 PM] Cec Madsen...................... says: Tasks for Scenario 1
(r-gma)
[6:39:08 PM] Cec Madsen...................... says: 1. create
publication_type('NetworkTCPThruput') - do this for 7 publications or 7
tables defined in scenario
[6:39:26 PM] Cec Madsen...................... says: 2. createPublisher
('site1') - a primary site
[6:44:32 PM] Cec Madsen...................... says:
publishersite1.associate(7 publication_types, policy information)
4. createSubscripiton(query=datafilter, consumerSetEPR)
note that conumseer is not explicitly modeled for this scenario (not a
consumer resource)
5. createPublisher('data center')
6. publisherdatacenter.associate(policy=3days) - we do reuse
publication_type
7. createSubscription()
[6:45:10 PM] Cec Madsen...................... says: Notes:
- publicaitontypes reused across (3) publishers
[6:47:19 PM] Cec Madsen...................... says: Items agreed upon
(done) so far:
1. createPublication instead of generic createINFODResource operations
2. createAssociation new operation with parameters: publisher,
publicationType, policy
3. getAssocation new operation (query associations)
4. dropAssociation
5. no updates of publicationType or publishers allowed (other than
association stuff)
6. publicationType is used to define schema vocabulary (doesn't include
policy anymore)
[6:48:07 PM] Cec Madsen...................... says: 7. no updates of
association allowed
[7:11:27 PM] Cec Madsen...................... says: Now going over each
task above providing detail parameter information:
[7:11:45 PM] malaikas says: Association s are immutable
In RGMA - Sites and and Data Centers are Publishers
In RGMA - NetworkTCPThruput is a PublicationType
In RGMA - NetworkTCPThruput table schema is a SchemaVocabulary avlue
[7:12:13 PM] Cec Madsen...................... says: 1.
createPublicationType('NetworkTCPThruput', 'sql92', 'create table
thruput...')
[7:12:40 PM] Cec Madsen...................... says: the
createPublicationType operation returns an EPR
[7:13:39 PM] malaikas says: In RGMA - Sites and and Data Centers are
Publishers
In RGMA - NetworkTCPThruput is a PublicationType
In RGMA - NetworkTCPThruput table schema is a SchemaVocabulary value
In RGMA - RetentionPeriod is a PolicyVocabulary value that characterizes
the assocaiation between a site (or data center) with the
NetworkTCPThruput
[7:18:24 PM] Cec Madsen...................... says: 2.
createPublisher('site1') - also returns an EPR
[7:18:30 PM] malaikas says: SchemaVocabulary -> Shape of the published
messages and how they can be qureied
[7:41:00 PM] malaikas says: Subscribers choose from associations when they
create subscriptions; Publishers choose from publication types when they
decide what to publish
[7:42:24 PM] Cec Madsen...................... says: 3.
PEPR.associate(PTEPR, RP=5mn, PRI=true) - do this 7 times, once for each
publication_type
with: publication_type_EPR = PTEPR and publisher EPR is PEPR, 'rgma',
'nmid source matches...' )
[7:44:28 PM] Cec Madsen...................... says: ignore 3. above -
overwrited as follows:
3. PEPR.associate(PTEPR, RP=5mn, 'rgma', 'incomplete, namid source
matches...')
with the last 2 parms being the vocabulary definition for 'description', a
new member of the association struct
[7:58:54 PM] Cec Madsen...................... says: add step between 3 and
4 to find the appropriate association (which will then be used to create
subscription)
[8:06:03 PM] Cec Madsen...................... says: Correction: step 3 is
against registry so rename to associate(..)
[8:07:09 PM] Cec Madsen...................... says: which still returns an
association epr, AEPR
[8:08:37 PM] Vijay Dialani says: Susan I have a visitor..i will have to go
[8:08:56 PM] malaikas says: ok - bye bye
have fun Vijay
[8:09:09 PM] Vijay Dialani says: Tomorrow at 9 am Chicago time?
[8:09:17 PM] malaikas says: Yes :)
[8:09:23 PM] Vijay Dialani says: ok...bye
[8:09:33 PM] malaikas says: maybe a bit after as I have to collect folk
[8:09:37 PM] malaikas says: byeeeeeeeeee
[8:09:38 PM] Vijay Dialani says: ok
[8:14:04 PM] Cec Madsen...................... says: ok so trying again:
3. AEPR = associate(PEPM, PTEPR, 'RP=5mn', 'rgma', 'incpomlete...')
4. createSubscription(PTEPR, policy, 'r=gma', 'select ...')
[8:14:21 PM] *** Vijay Dialani left this chat ***
[8:18:55 PM] Chris Kantarjiev says: a vocabulary looks like
struct infod_vocab {
char context[]; /* e.g., SQL92, RGMA */
char terms[];
};
[8:42:28 PM] Cec Madsen...................... says: so summary is that
typical grid scenario, ie r-gma, assumes that subscription does not want
to depend on an explicit association because may not know ahead of time -
which is an infod scenario somewhat involved (very different than typical
pubsub a la wsn)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.ogf.org/pipermail/infod-wg/attachments/20050704/0af914f4/attachment.htm
More information about the infod-wg
mailing list