Skip to end of metadata
Go to start of metadata

Main points/actions from discussion:

  1. Don's possibly concerns with RO folder structure were unfounded.
  2. Stian/Khalid/Sean to discuss release plan for RO model and changes
  3. There may be overlap on templates in RO Manager/myExperiment. Don/Graham to monitor.
  4. The notion of "blessed" or immutable annotations – for example provenance, creation, workflow type etc. What are they, and what are the mechanisms (templates?) that might be used to support this?
  5. KEGG services. SOAP services being replaced by REST. Heavily used in myExp, so provide a great opportunity for decay analysis. Propose a showcase running the workflows that use these services and collect provenance. The death of Biomoby also offers a (different kind of) opportunity. One is about repair, the other about preserving something that's dead. There is a discussion with Katy about this on Thursday – Jun will join.

57:17 David De Roure: Hi - I have a one-off project finance meeting at 3pm so won't be able to make the call. Carole is at an Elxir meeting in London.
02:19 also wont be able to make the call
03:05 Don Cruickshank: hi
03:07 Piotr Holubowicz: hi
03:23 Finn Bacall: hello
03:33 Sean Bechhofer: hi
04:00 Alan R Williams: Hello
04:07 Sean Bechhofer: Sean Bechhofer set topic to "3pm Combo Meet (29/10/12)"
04:09 khalid.belhajjame: hi
04:26 Don Cruickshank added Graham Klyne, Jun to this conversation
04:38 Jun: hello
04:55 Graham Klyne: Hi
05:09 Jun: is there an audio or just chat?
05:24 Sean Bechhofer: We usually do audio.
05:58 Jun: Can we (Graham and I) also be added to the audio please?(smile)
06:10 Sean Bechhofer: Plus some minutes/scribing.
06:19 Jun: Nice!
06:58 Robert Haines: I won't join audio today as I am firefighting biovel things but I will keep an eye on the chat if you need me
07:34 Sean Bechhofer: Ok. Thanks Rob. Would there be anything Biocat related to report or comment on?
07:57 Robert Haines: not this week - all quiet at the moment!
08:09 Sean Bechhofer: Tx.
08:34 Sean Bechhofer: I'll start audio in a moment.
08:40 Sean Bechhofer: Things to discuss
08:51 Sean Bechhofer: 1. wf4ever "myexperiment" showcases.
09:00 Sean Bechhofer: 2. server status
09:03 Sean Bechhofer: 3. AOB
09:19 Sean Bechhofer: Anything else to add?
10:34 khalid.belhajjame: Not sure if this should be added undermyExperiment. Katy mentioned in the last myGrid meeting that Kegg web services are moving from SOAP to REST, this will break a lot of workflows in myExperiment
10:50 Sean Bechhofer: Ok. Let's add that.
10:57 Sean Bechhofer: This message has been removed.
12:16 Sean Bechhofer: 1. myExperiment wf4ever showcases.
12:35 Alan R Williams: @Khalid - it will indeed. Katy, Rob and I are having a meeting about it on Thursday. Any representation from myE would be great. The skype chat is at 1100.
12:43 Sean Bechhofer: 83.
13:03 Sean Bechhofer: Piotr: Sprint finished tu/we but no time to add anything more.
13:08 Sean Bechhofer: ...somewhere in this week.
13:31 David De Roure: Re item 2 - the servers are racked according to the arrangement that was agreed, I met with Luke last week and today and he is sending account credentials to Dan and Don - the servers will be commissioned in the current rack then there'll be a relocation at some point, but Luke and Andy suggest this is best done after commissioning. Back to my meeting!
13:33 Sean Bechhofer: ...showcase involved more people.
13:44 Sean Bechhofer: ...would like to add full support for folders in RODL.
13:57 Sean Bechhofer: ...need to modify wf transformation service to support this.
14:27 Sean Bechhofer: graham: Nothing specific to add but wondering what migration path would be for those ROs not using folder structures.
14:33 Sean Bechhofer: ...will still be treated sensibly?
14:42 Sean Bechhofer: ...some discussion with Stian/Finn about folders.
15:05 Sean Bechhofer: ...all the web apps don't really have the concept of folders. May be paths in th URIs. but they're not treated as folders.
15:14 Sean Bechhofer: ...not much of a transition.
15:40 Sean Bechhofer: graham: was thinking of moving to folers. Add ORE constructs that represent the folders.
15:49 Sean Bechhofer: ...just want to know that we can migrate ROs painlessly.
16:09 Sean Bechhofer: Pitor: Once support is added. Just have to download RO and upload again. Go around the creation path.
16:16 Sean Bechhofer: Graham: Would prefer not to have to do that.
16:42 Sean Bechhofer: Sean: Issue with fodlers/don?
16:53 Sean Bechhofer: Don: Though folder structure would repalce existing view.
16:59 Sean Bechhofer: ...Concerned about finding things.
17:20 Sean Bechhofer: ...No longer an issue.
18:18 Sean Bechhofer: Sean: Are there records of discussion between Finn/Stian/Piotr?
18:28 Sean Bechhofer: Piotr: On the showcase wiki.
18:33 Graham Klyne: IIRC, the concern was that URIs would no longer directly reference aggregated resources. But that's not the case: URIs always identify independently of folder context.
18:39 Sean Bechhofer: ...Some discussion last week about finding root folders.
18:49 Sean Bechhofer: Finn: Not on the wiki, but could be,
18:58 Sean Bechhofer: Piotr: Some changes in the RO model?
19:02 Sean Bechhofer: ...or just the examples.
19:24 Sean Bechhofer: Stian: Haven't changed the documentation yet.
19:30 Sean Bechhofer: ...awaiting the next version.
19:44 Sean Bechhofer: Sean: Schedule for the next version?
20:18 Sean Bechhofer: ACTION: Stian/Khalid/Sean to determine a "release" plan for the RO model.
20:34 Sean Bechhofer: Sean: Other things feeding in?
20:56 Sean Bechhofer: 84. Improving the RO interface
21:10 Sean Bechhofer: Don: Getting the annotations up to scratch ready to work on this.
21:28 Sean Bechhofer: ...Can point myexp at an RO and get to its physical layouts and view resources + annotations.
21:36 Sean Bechhofer: ...Can add annotations and replace them.
21:45 Sean Bechhofer: ...simple literals, title/description etc.
21:55 Sean Bechhofer: ...WIll delete exisitng annotations of th same strcuture.
22:15 Sean Bechhofer: ...should more compliacted ROs be created. Will have a debug view for looking at the annotations directly.
22:27 Sean Bechhofer: Sean: debug view and expert view?
22:46 Sean Bechhofer: Don: Yes. Looking good so far. Need to add template support so that people can add stuff to fill in.
22:59 Graham Klyne: Interesting: template support is in my sights for RO manager too (smile)
23:04 Sean Bechhofer: ...nothing with folder views as yet.
23:14 Sean Bechhofer: ...mockups not really there yet.
23:20 Sean Bechhofer: Finn:: Needs folder stuff to work properly.
23:44 Sean Bechhofer: graham: Comment more out of interest.
23:58 Sean Bechhofer: ...not an obvious interaction. At the end of the day exchange is RF annotations.
24:34 Sean Bechhofer: graham: Trying to align things might introduce more complication.
24:38 Sean Bechhofer: don: +1
24:49 Sean Bechhofer: Graham: Quickest to do our own thing and then potentially abstract later.
24:59 Sean Bechhofer: Sean: Keep an eye on it.
25:18 Alan R Williams: I am on the call
25:42 Sean Bechhofer: Sean: How does this play with the roundtripping/stickiness etc.
25:53 Graham Klyne: IIRC, the perceived/experienced problem was with annotations created in Taverna.
25:54 Sean Bechhofer: Don: REST API gives access to original artefact.
26:13 Sean Bechhofer: ...difficulty roundtripping is to do with the t2flow being downloaded, which we're not changed.
26:32 Sean Bechhofer: ...THouht that Alan envisaged we'd move bundles (worklfow + annotatiosn) around.
26:40 Sean Bechhofer: Alan: only sort of. (smile)
26:57 Sean Bechhofer: ...basic idea was to be able to pull in annotations from known soruces. So if annotation done on myexp.
27:03 Sean Bechhofer: Don: Would that come from API?
27:05 Graham Klyne: I understood that the near-term plan was to extract but "lock" taverna annotations; longer term to have taverna use RO model too.
27:33 Sean Bechhofer: Alan: More imaging that given the idea of the workflow you'd be able to query myexp and get the annotations on that workflow (or any other objects in the workflow).
27:57 Sean Bechhofer: Sean: These need to be reconclied?
28:04 Sean Bechhofer: Alan: Or shown to the user and reconclided.
28:23 Sean Bechhofer: ...Question as to whether myexp should be allowed to create those annotations.
29:03 Sean Bechhofer: ...For example if I've created something it wuld be annoying if something else could change the creator.
29:18 Sean Bechhofer: Sean: DO we know what those immutable annotations are?
29:22 Sean Bechhofer: Alan: No
29:28 Sean Bechhofer: Sean: Do we need to know?
29:45 Graham Klyne: Just do it?
29:47 Sean Bechhofer: would we go about finding what they are.
30:29 Graham Klyne: "an exercise for the implementer", methinks.
30:38 Sean Bechhofer: There needs to be a mechanism that specifies annotations that shouldn't be messed with (e.g. workflow type).
30:42 Jun: and some of the provenance information?(smile), like the creator, creation date
30:50 Sean Bechhofer: How those annotatinos are chosen is up for grabs.
31:07 Sean Bechhofer: ...potential impact for the UI.
31:11 Sean Bechhofer: Don: Role of templates/
31:18 Jun: +1 to templates
31:39 Graham Klyne: Ah, I think we have slightly different notion of templates (smile)
31:39 Sean Bechhofer: Tempaltes wouldn't offer immutable stuff. Protection via obfuscation.
32:39 Sean Bechhofer: Piotr: When you talke about RODL doing annotation. From my perspective some thigns are managed by RODL and some are annotations. So different levels of credibility. Can't directly change manifst in RODL. Evolution etc also different.
33:16 Sean Bechhofer: Sean: "Blessed" annotations that shoildnt be messed with.
33:25 Sean Bechhofer: Pitor: Some things are facts not annotations.
33:27 Graham Klyne: (They're all annotations to me (wink) )
33:53 Jun: +1
34:08 Sean Bechhofer: All annotations are equal, some are more equal than others (smile)
34:12 Graham Klyne: I think we're slightly confusing model with implementation in this discussion.
34:24 Piotr Holubowicz: I just don't take the manifest as an annotation, unlike the RO manager (wink)
34:43 Sean Bechhofer: Graham: some of this detailed discussion is in implementation territory.
35:44 Sean Bechhofer: Two things out of this discussion:
36:02 Sean Bechhofer: 1/ Templates – Graham/Don to have a view on this.
36:14 Sean Bechhofer: Graham: Already arranged for standups about this.
36:27 Sean Bechhofer: for such disciussions.
36:38 Sean Bechhofer: 2/ Understanding "blessed" annotations.
37:22 Sean Bechhofer: Graham: May want to surface some notion of mutability in the model.
37:53 Sean Bechhofer: ...perhaps not immediately. But potentially some discussion between implementors and modellers.
38:05 Sean Bechhofer: ...model as an abstraction of the RO structure.
38:22 Sean Bechhofer: ...implementation as a realisation of the model.
38:32 Graham Klyne: = RO model; RODL, RO manager provide implementations
38:50 Sean Bechhofer: 85.
39:09 Sean Bechhofer: Piotr: Little to say here. Just today some benchmarking on sandbox.
39:18 Sean Bechhofer: ...Surprisingly slow.
39:37 Sean Bechhofer: ...All the servers on this cluster have network storage and similar writing speeds.
39:43 Sean Bechhofer: Don: Need to get it running locally.
39:46 Graham Klyne: "piotr are you running in VM? Maybe short of RAM. Network storage should be fast if there's enough RAM.
39:59 Sean Bechhofer: Piotr: Not the only reason. Some programming issues.
40:10 Sean Bechhofer: Around 15 MBits/sec. Laptop has around 100.
40:19 Sean Bechhofer: Bound to see a difference comparing locally.
40:51 Sean Bechhofer: Don: Decided just to hit RODL every time I need it. Getting things right rather than worrying about speed.
41:01 Sean Bechhofer: ...What about the deadline for the end of this month?
41:25 Sean Bechhofer: Piotr: Not presenting enough to be rally concerned about speed. First get it working, then worry.
41:40 Sean Bechhofer: Pitor: 4Gig on the machine. Not a problem of lack of memory.
41:58 Sean Bechhofer: ...Quicker RODL in a week or two. Need to move some content around.
42:16 Graham Klyne: @don - I think that's right (just hit RODL) - HTTP caching should provide some performance boost (for access).
42:30 Piotr Holubowicz: @Graham +1 RODL supports cacheing
42:41 Sean Bechhofer: 86.
42:49 Sean Bechhofer: Piotr: Nothing to report.
42:58 khalid.belhajjame: I know that Reinout was in communication with Stuart
43:41 Sean Bechhofer: Khalid: KEGG webservices used heavily in myexp. SOAP services are still provided, but they're being replaced by REST services.
44:07 Jun: It IS an opportunity!
44:10 khalid.belhajjame: Here is the kegg rest api
44:11 khalid.belhajjame:
44:12 Sean Bechhofer: Sean: Seems like a real opportunity!
44:28 Don Cruickshank: (sorry - i need to drop off the call now)
44:42 Sean Bechhofer: Khalid: Page explains the REST API. The names of operations change and the arguments. Trying to provide some generic operations.
44:47 Sean Bechhofer: <tx Don>
44:50 Graham Klyne: @don, @piotr ... if that's slow, I think we want to avoid building more cacheing into myExp. (If necessary use cache middleware?) Updates more difficult, but how much of the overall problem are these. Maybe look into SWORD 2 work which provides for background transfer of bulk data?
45:14 Sean Bechhofer: Some detective work needed to see which operation would map to which URI.
45:34 Sean Bechhofer: Alan: It isn;'t just an easy flick of a switch. Looks like a total reimp[lementation.
45:50 Jun: We should run all the workflows using kegg and record their provenance now!
46:00 Sean Bechhofer: Jun+1
46:11 Sean Bechhofer: Khalid: From the end of the year.
46:23 Sean Bechhofer: Jun: Least we could do is run all the workflows and keep a copy of their provenance.
46:36 Sean Bechhofer: Sean: And do that as ROs!
46:45 Graham Klyne: This sounds like an opportunity for automatedb repair: provide REST interfaces that can be swapped in, and the auto-repair should be able to use these when needed?
46:48 Sean Bechhofer: Khalid: Provenance corpus has these workflows.
47:00 Sean Bechhofer: Jun: SHould be as complete as possible.
47:20 Sean Bechhofer: Khalid: Is there any formula we can find to guess the REST URls.
47:37 Sean Bechhofer: Alan: Need to analyse first which of the SOAP operations are used b the wf on myexp.
47:52 Graham Klyne: I think automated interface rewriting is a big ask. How many are there to be updated?
48:02 Sean Bechhofer: Jun: That's a static level mapping. Provenance supports dynamic mapping.
48:38 Sean Bechhofer: Sean: One or two showcases?
48:52 Sean Bechhofer: Khalid: Finding wfs is a SPARQL query.
49:07 Sean Bechhofer: ACTION: Generate showcase for this.
49:26 Sean Bechhofer: Khalid: Could do with help from Katy.
49:47 Sean Bechhofer: Alan: Discussion with Katy on Thursday. Also SOAPLAB services disappearing. And biomoby is dying.
50:10 Sean Bechhofer: Graham: Isn't this dealt with by subworkflows? Need to provide new subwf?
50:31 Sean Bechhofer: ...Components?
50:42 Jun: Would be interested to know whether anyone has the right to mirror the host of these services
50:43 Sean Bechhofer: Alan: No, it's usually a single SOAP call.
51:06 Sean Bechhofer: Graham: Does Taverna allow you to swap out a SOAP call to a subworkflow?
51:18 Sean Bechhofer: ...into automated wf repair.
51:55 Sean Bechhofer: Sean: Would all these servce issues be dealt with in similar ways?
52:13 Sean Bechhofer: Alan: Soaplab/Kegg similar. Biomoby more problematic. No replacement services.
52:26 Jun: @graham, That's why provenance is potentially invaluable, for repair and replay
52:50 Sean Bechhofer: Sean: So different issues. One about repair, the other about preserving something that's dead.
52:59 Piotr Holubowicz: (I need to leave in 5 mins)
53:05 Graham Klyne: (Yeah. Also, with REST interfaces, web archive systems might be useful in some cases - e.g. BioMoby registries?)
53:29 Sean Bechhofer: SOaplab and Biomoby also candidates for gathering provenance corpus.
53:43 Jun: Can I join the Thursday's call? (smile) please?
53:43 Sean Bechhofer: Alan: Katy knows how to fix some of these.
54:29 Sean Bechhofer: ACTION: Jun to join Skype chat on Thursday at 11:00 (UK or CET(smile)
55:42 Alan R Williams: It is 1100 UK time
55:55 Sean Bechhofer: AOB?
56:43 Sean Bechhofer: We're done.

  • No labels