Skip to topic | Skip to bottom

Provenance Challenge

Challenge
Challenge.FirstOPMWorkshopChatLog

Start of topic | Skip to actions
from paul Groth to All Participants:
hi 
from Jim Myers to All Participants:
hi paul
from Jim Myers to All Participants:
if I give you control, can you share the app you're taking ntoes in?
from paul Groth to All Participants:
it's just a text file 
from paul Groth to All Participants:
so sure 
from Jim Myers to All Participants:
I'm hoping to get a bit more info on what's being discussed and perhaps to chime in via text 
from Jim Myers to All Participants:
OK - you sould be able to share app or doc now
from paul Groth to All Participants:
  do you see it 
from paul Groth to All Participants:
? 
from Jim Myers to All Participants:
yep
from paul Groth to All Participants:
how do I update? 
from Jim Myers to All Participants:
You just type I think -
from Jim Myers to All Participants:
did you share your live doc or upload?
from paul Groth to All Participants:
so i think that's better 
from Jim Myers to All Participants:
I think you can share the app/doc directly rather than your desktop (which I think you did)
from Jim Myers to All Participants:
but both work
from Jim Myers to All Participants:
was a consensus on agents being an important concept reached?
from paul Groth to All Participants:
the consesus was that the concept of wascontrolledby 
from paul Groth to All Participants:
was important but not the concept of agent 
from Jim Myers to All Participants:
I disagree of course - agents are the only thing in the model that can change state
from paul Groth to All Participants:
can't processes? 
from Jim Myers to All Participants:
processes have state?
from paul Groth to All Participants:
processes execute 
from paul Groth to All Participants:
maybe you can make a case? 
from Jim Myers to All Participants:
yes - they take things in one state and prodcue other things (which we may think of as the same things as inputs in a different state
from paul Groth to All Participants:
did you want to talk? 
from Jim Myers to All Participants:
I'm hearing so little it is hard to know what to say
from Jim Myers to All Participants:
the alternate relationship is transitive?
from Jim Myers to All Participants:
a alt b, b alt c, --> a alt c
from paul Groth to All Participants:
it is transitive 
from Jim Myers to All Participants:
I think of witnesses to a traffic accident - they will all claim to give true accounts and some may consider their accounts as being alternates to others, but a judege will really have to decide who is telling the truth and which accounts are really alternates
from paul Groth to All Participants:
ok  
from paul Groth to All Participants:
that's similiar to what luc 
from Jim Myers to All Participants:
we don't have the concept of judge, but it is implicitly a part of aggregating accounts
from paul Groth to All Participants:
says 
from paul Groth to All Participants:
yeah 
from Jim Myers to All Participants:
OK - one thing we can do is just capture the chat log at the end as part of the notes
from paul Groth to All Participants:
sure 
from Jim Myers to All Participants:
I think we put in alternates to assist te judge - allowing witnesses to say tey tink it is an alternate helps te judge
from Jim Myers to All Participants:
we have few/no rules as to how one combines accounts
from Jim Myers to All Participants:
if one account claims alternates, it is transitive, if the assertions are in diferent accounts, wecannot assume transitivitiy - we can't even assume that both are truthful
from paul Groth to All Participants:
exactly 
from paul Groth to All Participants:
well I don't think we can assume any assertions in the model are true 
from Jim Myers to All Participants:
legal alternates begin/end on the same artifacts, i.e. you don't have alterantes if one describes additional processing
from Jim Myers to All Participants:
this is why we added sub accounts
from Jim Myers to All Participants:
(one reason anyway)
from Jim Myers to All Participants:
Since accounts can be incomplete, the idea of alternates was introduced to be an asserion that, even if, for example, a says there was one output and account b says there were two, someone says they are about the same process by saying their alternates
from Jim Myers to All Participants:
and hence a judge can assume a just didn't see everyting that happened.
from Jim Myers to All Participants:
I think we have to question what the value of the alternate alternate definitions are
from Jim Myers to All Participants:
the current one elps a judge, simple overlap types of definition don't seem to help very much...
from paul Groth to All Participants:
we want it to be explict 
from Jim Myers to All Participants:
(I cant assume transitivity, etc
from Jim Myers to All Participants:
what to be explicit?
from Jim Myers to All Participants:
the definition?
from paul Groth to All Participants:
the notion of alternates e.g. overlaps 
from paul Groth to All Participants:
that ou have to explicitly document it  
from Jim Myers to All Participants:
vocabulary for what?
from paul Groth to All Participants:
trying to type it  
from paul Groth to All Participants:
or remember what they meant :-) 
from Jim Myers to All Participants:
we had thought to start identifying useful ontologies to use for data, people, etc. rather than try to pull those into OPM - provenance is not the only driver for stnadardizing those descriptions, so we're te wrong set of people to standardize them.
from paul Groth to All Participants:
exactly 
from Jim Myers to All Participants:
we could suggest some to use in the challenge peraps...
from paul Groth to All Participants:
sure 
from paul Groth to All Participants:
are you ready to talk? 
from Jim Myers to All Participants:
Can we start thinking about capturing the proposed canges and the rationale for each for comment BEFORE revising the doc?
from Jim Myers to All Participants:
potential challenge 3 ideas:
from Jim Myers to All Participants:
nail down the ML and RDF level descriptions
from Jim Myers to All Participants:
understand how, if all theh entrants were witnesses producing different accounts, could we get to the 'real' combined account?
from Jim Myers to All Participants:
strive for interoperable services rather than interoperability via exported docs
from Jim Myers to All Participants:
(not sure this is doable given that we don't have syntaxes yet or common identifiers for data)
from Jim Myers to All Participants:
anoter option - ask for extensions to the challenge 2 workflow - show how some people can deal with non-repudiation/proof, how some can embed the workflow in human/financial/other processes/etc
from Jim Myers to All Participants:
another option - how do we work with lots of provenance - interfaces, queries, etc. to understand undreds/thousands/millions of runs
from Jim Myers to All Participants:
re: extensions (2 comments back) - how can people link between scientific/mathematical/computational provenance
from Jim Myers to All Participants:
representation of streams/versions/etc migt be interesting...
from Jim Myers to All Participants:
collections (per the phone discussion) is another
from Jim Myers to All Participants:
Here's a challenge  - ave people submit a set of prov trails and try to derive the underlying workflows - can we recognize were if statements, loops, etc. were
from paul Groth to All Participants:
that's a challenge 
from Jim Myers to All Participants:
:-)
from paul Groth to All Participants:
that's pretty hard to do even if you understand the underlying model 
from Jim Myers to All Participants:
are we ready for performance related challenges - fastest, most scalable, multiuser/distributed, etc. If the challenge is multidimensional this way, we probably could avoid the issue of one 'best' system and focus more on best for x or y or z use case...
from Jim Myers to All Participants:
XML to RDF sould just be a standard XSLT, not a challenge...
from Jim Myers to All Participants:
another thought - make the challenge about pathological workflows - very large, very complex, involving human processing, etc. - challenging people to represent them in OPM or coherent extensions 
from Jim Myers to All Participants:
wy do we need components? IN prior challenges, people just mocked up a process that read a file and produced a file...
from Jim Myers to All Participants:
go Luc!
from paul Groth to All Participants:
:-) 
from Jim Myers to All Participants:
Let's capture the provenance of a traffic accident!
from Jim Myers to All Participants:
doesn't #9 have to come before #6?
from paul Groth to All Participants:
i don't think so 
from Jim Myers to All Participants:
we don't know what the challenge's goals are but we know what workflows will help us get there? or?
from paul Groth to All Participants:
are goal is interoperabilty through opm, i think 
from Jim Myers to All Participants:
or is 9 just te details of when, etc.
from Jim Myers to All Participants:
so teh challenge is - all people generate OPM for all workflows selected?
from Jim Myers to All Participants:
or we read/write each other's OPM or?
from paul Groth to All Participants:
yeah, we read/write each other's OPm 
from Jim Myers to All Participants:
so challenge 2 again with the hope that additional workflows (reviewed for what attributes?) expose otehr issues? 
from Jim Myers to All Participants:
NSF likes to put grant deadlines around Christmas - can we sift to end of January?
from paul Groth to All Participants:
just shout 
from Jim Myers to All Participants:
most sophisticated, most useful, most mature, most scalable, most ... categories for prizes?
from Jim Myers to All Participants:
arrgghh!
from paul Groth to All Participants:
oh well 
from paul Groth to All Participants:
i'll copy the comments. 
from Jim Myers to All Participants:
thanks for translating for me, a poor text bound lad...
from paul Groth to All Participants:
no worries 

-- PaulGroth - 20 Jun 2008
to top


You are here: Challenge > OPM > OpenProvenanceModelWorkshop > FirstOPMWorkshopMinutes > FirstOPMWorkshopChatLog

to top

Copyright © 1999-2012 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback