Information Dissemination BOF (INFOD-WG) ======================================== Susan Malaika, IBM motivation: replication, third party delivery and sub-patterns goals: define operations and interfaces necessary to support information dissemination define realizations for typical scenarios to hide the complexity of the wide-range of infod options position the infod model within ws-notification scenarios: A: spam pattern: pushing stuff to consumer directly B: pub sub pattern: publisher publishes to a broker, consumer receives material on demand C: sub pub pattern: consumer asks for stuff and publisher publishes on demand according to subscriptions Q: will we address all scenarios - there is a possible overlap with other groups A: to be defined Q: will use cases be sufficient to drive us A: we have quite a lot from the DAIS work Dieter Gawlick, Oracle ---------------------- pub-sub scenarios: - send a message to me and then try to call me over phone - send message to bill or call information - send message to all car dealers within 10 miles who have an interesting rate for a given model - spam, pub/sub combination? Q: late binding? A: yes reaching any mix of consumers publishing relevant info - publishers have to capture the right info. there is no limit on what *could* be of interest - base publication on subscription requests! demand driven publication - this function complements the pull-data access business scenarios - replication: replicate selected info to selected recipients as it becomes available - EAL workflow - Business activity monitoring - Radio Frequency identifyers basic models - elements: publisher, consumer. publisher ---comm channel---> consumer extended model - add registry - publisher, consumer and registry of profiles - reg. for publishers: name, what info is available, language to specify interest - reg. for subscribers: name, when and what is of interest, what should be done if info is available operational characteristics - # messages per second - # of subscribers - security - transactions Q: you assume that all consumers are registered - can this be avoided A: Yes Q: How can you do it without a trigger mechanism (i.e. acount going below some level) A: This is a model - need suitable implementation Q: It would be to be able to be able to define in terms of composition of services A: This sounds like a good direction - but not there yet Q: how does it relate to Grids? A: it is an extension of DAIS when timeliness is important Q: Are there a lot of security implications A: yes Q: this is powerful - but will it scale down well. Can we cope with the overheads? Perhaps turn off the slow bits e.g. avoid late binding as necessary A: We are doign it now Q: is this WG or RG - there is a lot of research needed A: WG Q: what is diff between data movement and dissemination A: dissemination is an informed movement of data Dave Pearson ------------ DAIS presentation - relation to INFOD - Requirements identified need to deliver data to and from 3rd parties x event notifications x distributed/federated queries x immediate and deferred result set delivery - data movement out of scope - info based discovery in science needs x topic based sub x publishing relevant info - business scenarios have equivalents - timely delivery of info Ann Chervenak ------------- OREP presentation - relation to INFOD Ann proposes to get rid of OREP and go under the INFOD umbrella Various opinions are expressed, for and against Steve Fisher ------------ RGMA presents RGMA features etc - difference to suggested INFOD path: registry is hidden, in INFOD it's exposed. - use relational model explicitly - only relation to others is logging. - INFOD should be active in the relation to OGSA Discussion of Charter and next steps ------------------------------------ relation to wsrf, ws-notification question is how narrow/wide you want to have the charter. charter: define interfaces necessary to support INFOD define realizations for typical scenarios aim is to define a type of pub-sub? strong collision with 4 other groups! logging, ws-notification etc. everyone is trying to be a fundamental mechanism.. work to be done in description work to be done for positioning wrt other groups should abstract out of delivery system and other systems the notion of matching publishers and subscribers which can be composed of other systems Q: Are the collisions with other working groups avoidable A: Overlaps are a real concern - but until some work is done can we know? There was general agreement that this is a useful activity with good support from the community. An announcement will go to the data area and dais list when SourceForge goes ahead.