Table of Contents

Time/Place

This meeting is a hybrid teleconference and IRC chat. Anyone is welcome to join...here's the info:

Agenda

Quick review of High Level Storage

  • Rationale
  • Status

Roadmap (create in JIRA?) for HLStorage. Potential major topic areas:

  • Review fedora object model, propose revised DigitalObject interface
    • Versioning
  • Design and comment on HLStorage APIs
    • Return values
    • Transaction support?: What can or need we do now, and how
      can/will/must it influence our APIs?
    • Alternatives: Compensation, Reliable Designs, Idempotent Operations, Eventual Consistency
    • Cross Cutting Topics: ACID, Synchronous vs Asychronous operations
    • Defining what "storage" means and deriving a common nomenclature
  • HLStorage internal architecture
    • Layering/chaining
    • Data-oriented messaging
    • Indexing

How to proceed

  • Special topic meetings or special interest groups with periodic report-back?
  • Use fcrepo-dev list, or would that generate too much traffic? Different people may be interested in different aspects.

Audio/Chat Log

[11:10] <cwilper> Meeting now: https://wiki.duraspace.org/display/FCREPO/2010-09-28+-+Special+Topic+-+Moving+Forward+on+High+Level+Storage
[11:11] <cwilper> Aaron giving an update on HLStorage status
[11:11] <cwilper> "Time to kick off concrete design activities"
[11:11] <cwilper> "Next few months = heavily design/theory oriented"
[11:12] <cwilper> Dan: "forward-looking roadmap but breaking into pieces"
[11:13] <cwilper> Aaron: "Scoping for Fedora 4.0, can be less conservative about changes"
[11:15] <cwilper> Aaron: "3 major threads of discussion moving forward: 1) Fedora Object model itself (what changes need to be made in order to facilitate this?  DS Versioning?"
[11:17] <cwilper> "2) API itself. Have initial proposals. First from Aaron was simple, Asger introduced idea of separating into R/W interfaces.  Have discussed return values (ticket id for reflecting on status on async storage)."
[11:17] <cwilper> "3) Architecture.  Do we go with layering/chaining? Data-oriented messaging, caching, what other activities can happen in the HLStorage layer"
[11:20] <cwilper> "Suggest representing the HLStorage roadmap into components"
[11:21] <cwilper> Chris: "Could use Components in JIRA, not sure about sub-components, but we could fake them"
[11:28] <cwilper> Aaron: "Should we have a Component for changes to Fedora that may be necessary"
[11:28] <cwilper> Chris; "Important to identify when that's the case"
[11:29] <cwilper> Dan: "Design needs to be informed by things like, how Fedora is used as a storage resource itself"
[11:30] <cwilper> Chris: "Does that mean 'effect on Fedora' category makes sense"
[11:30] <cwilper> Dan: "Yes."
[11:31] <cwilper> Chris: "Effect on Fedora? as category name"
[11:33] <cwilper> Dan: "Suggest using Wiki for design-oriented tasks prior to getting to rigid with JIRA"
[11:33] <cwilper> Aaron: "Let's switch to -How to proceeed-"
[11:34] <cwilper> "Use Future Special Topics mtgs for this, or Special interest groups for this"?
[11:34] * mdiggory wants to toss out a tangent.... https://wiki.duraspace.org/display/FCREPO/High+Level+Storage?focusedCommentId=22024650#comment-22024650
[11:36] <cwilper> Thanks Mark...interesting to consider.
[11:37] <cwilper> Chris: "I like the idea of splitting off into interest groups for reasons Aaron mentioned"
[11:38] <cwilper> "(Timing and the fact that it might dominate the committer mtgs)"
[11:39] <cwilper> Aaron: "Communication on dev list? Concern: too much traffic"
[11:39] <cwilper> Chris: "Traffic=good"
[11:41] <cwilper> Dan: "+1 on keeping discussion on the list.  People could filter"
[11:41] <cwilper> Steve: "+1 on dev list"
[11:42] <cwilper> Asger: "Lots of opinions on design and not a lot on how we carry out the design"
[11:42] <eddies> (afk)
[11:42] <cwilper> s/carry out the design/carry out the discussion/g
[11:43] <cwilper> Aaron: "Fedora Object Model is a good area to start with.  Let's identify the sub-issues."
[11:44] <cwilper> "In general, make sure we have the items listed under each topic"
[12:01] <cwilper> (brainstorming sub-areas under "Fedora Object Model")
[12:02] <cwilper> So far:
[12:02] <cwilper> * Versioning
[12:02] <cwilper> * Audit Trail
[12:02] <cwilper> * Separating the kind of Content (XML vs non-XML) from where you store it ("Inline XML?")
[12:03] <cwilper> * Extensibility of Object Properties vs just using RELS-EXT/INT
[12:04] <cwilper> * What convenience methods, if any, should be in the new DigitalObject interface for things we normally categorize as "semantic" (e.g. the hasModel relation?)
[12:04] <Asger> Suggestion: Content Model Semantics
[12:05] <Asger> Ie. semantics that allows applications to reflect on there stuff goes in the object
[12:13] <cwilper> Aaron: "Reading mdiggory's post...let's step back and see the thinking and see what's applicable.  Want to understand conclusions."
[12:14] <mdiggory> I wish I could be in the call, but I'm stuck on the phone now... will need to followup afterward... sorry...
[12:16] <cwilper> Ok, Mark, it sounds like it'd be useful to have you in on (at least) a future discussion where we talk about the DigitalObject interface.
[12:19] <cwilper> Aaron: "Who's interested in this particular sub-topic. (Chris, Asger, probably more)..I'm inclined to put this as an item under the Fedora Object Model thread of interest"
[12:19] <cwilper> (continuing with list):
[12:20] <sbayliss> I have to leave right now unfortunately, looking forward to discussing further on the list etc
[12:20] <cwilper> * Understand DSpace2 modeling decisions and how they apply to the new Fedora DigitalObject thinking (see if mdiggory can be on a call sometime soon to help us understand this a bit better)
[12:22] <sbayliss> (I have posted to the committers list re the up-coming release status)
[12:25] <cwilper> Eddie: "caught tail end of discussion re:diggory's post.  i'm interested in whole-object versioning and required relationships in a content model"
[12:26] <Asger> My summary of Eddie: Regarding a set of Fedora Objects as a Fedora Record, or some sort of combined resourse
[12:27] <Asger> We have actually worked on that issue
[12:28] <Asger> In Enhanced Content Model
[12:42] <eddies> (must run)
#trackbackRdf ($trackbackUtils.getContentIdentifier($page) $page.title $trackbackUtils.getPingUrl($page))
  • No labels