DECADE WG Minutes Meeting: IETF-80, Monday, 13:00 -- 15:00, Prague, 2011-03-28, Location: Grand Ballroom, Hilton Hotel Chairs: Richard Woundy, Haibin Song Note taker: Dirk Kutscher , edited by Haibin Song and Rich Woundy [These minutes represent a condensed summary of the meeting. Fully detailed discussions can be heard on the audio recording of the meeting, currently found at http://ietf80streaming.dnsalias.net/ietf80/ietf80-ch5-mon-pm.mp3, starting at offset 00:31:45.] Rich Woundy (RW) -- Agenda, Chair slides Introduction to DECADE. Any questions or any refinement to the agenda? Rich mentioned that Lucy Yong has a conflict session will be here later, so her presentation is moved after the secure naming presentation. Introduction to the agenda. Two documents have finished WGLC, and two documents will be on next and follow the same procedure of problem statement and survey (select three reviewers, send comments to the list, make another version of the draft, start WGLC). Will start re-chartering discussion. No comments. ---------------------------------------------------------------------- [audio offset 37:05] Haibin Song -- Problem Statement Haibin presenting Resolved issues addressed all comments now informational (instead of PS) added references and more Asking for further comments before submitting to IESG review no comments ---------------------------------------------------------------------- [audio offset 39:00] Akbar Rahman -- Survey Akbar presenting Summary of changes -01 reviewed at IETF-79 received a couple of comments reviewed by chairs WGLC more comments now: -04 with all comments addressed About 20 significant comments. Categorization, access control, etc. Add new survey of information centric network. Open Issues nothing aware of Asking for any additional issues no comments Rich Woundy: at this point no further cycles of problem statement and survey -- these will be with IESG soon ---------------------------------------------------------------------- [audio offset 42:50] Richard Alimi (RA) -- DECADE Architecture RA presenting Architecture Objectives (Reminder) Changes since previous version now WG work item added details on data sequencing, naming, DECADE protocols etc. Open Issues (1) Data Sequencing and Naming naming to based on content hashes app names to be independent of DECADE names Richard asking for WG comments Question Richard Woundy: when doing names based on hashes -- are we providing unique names? RA: unique names could be an option -- Open Issues (2) DECADE protocols (DRP and SDT) -- Open Issues (3) different approaches for DRP and SDT. Approach 1: SDT carries DRP tokens in line. Approach 2: Separate protocol. Pros and cons of two approaches Question Richard Woundy: how many people have read draft? reasonable number of people RW: let's ask for consensus for the proposed way forward for the issues RW: open issue 2 -- any objection against SDT/DRP logical seperation No objection RW: open issue 3 -- what option do you prefer? RA: some prefer option 1 - client is responsible for delegating tokens for requests - has some benefits (RA describing) comment by Akbar: also prefer approach 1 RW: option 1 would mean that we need some extension mechanism to carry token, right? RA: yes, right RA: does NFS have a field where we carry arbitrary additional parameters Dave Harrington: I have no clue. But NFS WG may have Spencer N: you could implement a RPC mechanism, would be heavy-weight there is an additional API feature question David McDysan: on issue 2 -- SDT could use anothing existing protocol, does not have to be an IETF protocol Richward Woundy: according to charter, we have to leverage existing IETF protocols if possible -- Open Issues (4): Querying Server Status 1) add status requests to existing protocols 2) allow SDT responses to include DRP metadata RW: any comments, do co-authors agree? The co-authors nodded. Spencer: it may not be easy to provide this status information, e.g., in NFS RA: this is not meant for overall status information -- just for resource information relevant for single client Spencer: but it may not even available for a single client RW: if we go with option 2 and include some DRP meatadata -- is it the case that we want to have unified status? What if DRP says OK, but NFS side not? RA: don't have enough implementation experience -- could be that status info is different -- might have to handle that -- don't have a good answer yet. RW: Any other comments? No. -- Next Steps on TODO List RW: how many more draft revision do you think we need RA: probably 1.5 RW: maybe let's have one and then have it reviewd RW: volunteers for review? Börje Ohlman Dave McDysan Ove Strandberg ---------------------------------------------------------------------- [audio offset 69:15] Richard Alimi -- DECADE Requirements Change Summary (1) -- Change Summary (2) some new requirements -- Open Issues question Akbar: since charter explicitly talks about P2P, why now P2P-agnostic? RA: for normative text, we don't need to state something about whether P2P or not. Better to be explicit of properties Akbar: OK, I see the logic, but regarding architectures: should it also be P2P agnostic RA: fair criticism. for use cases, mentioning P2P seems OK. when it come to protocol design, preference is to be more explicit about actual properties RA: proposal: in normative text we don't refer to P2P Akbar: don't have an answer, but was good feedback RW: we said "P2P and similar" David McDysan: agree to that recommendation. Would like to propose to even extend it to cover P2PSP as well, DECADE could be used in those contexts RW: any other comments? no -- Open Issues (2) should properties of a naming scheme be added to requirements? next revision would have some explicit statements on naming schemes, would that be reasonable? Dave Harrington: I would be careful about specifying requirements because one of the main things the WG is supposed to do it to see whether existing protocols can suffice. Better make sure that we have features that are actually needed. Börje Ohlman: having a good scheme would be essential for ensuring that other apps can use DECADE, of course making sure that we do not exlude existing protocols RA: OK to say that we will be very careful? Dave Harrington: should be careful to distinguish between requirements and nice-to-haves. And make sure naming scheme supports different apps Aaron Falk: you should be aware that there are several research projects, where naming is a fundamental properties. Not suggesting that you build that in, but talking to folks is recommended. Dirk: will take care. -- Open Issue (3) on server information RA: should server status be extended? recommending to keep definition narrow, providing limited status no comment RW: do co-authors agree? yes (David Bryan is not here) RW: no-one disagrees RW: who has read draft a few people -- Items to be addressed requirement of a discovery mechanism, not planning to create a new one RW: document maturity: have to resolve naming and resource disovery how many new versions? RA: after issues are resolved, then ready volunteers for reviewing: Akbar Martin Stiemerling Dave McDyson RA: do we keep requirements doc open? RW: have to get it done, don't leave it open ---------------------------------------------------------------------- [audio offset 92:00] Börje Ohlmann: Secure naming structure and P2P app interaction Börje presenting General motivation -- Main message of this draft: common, app-independent naming scheme for information objects is needed for Internet -- requirements for naming scheme question RW: is uniqueness a property? Börje: if you use it for hashing, uniqueness is statistical RW: DECADE's primary interest would be uniqueness? Are others looking for more (PPSP)? Börje: see next slides RW: any other questions? no -- Secure naming for P2P -- Secure naming in DECADE -- Secure naming for CDNs -- Draft changes 01 -> 02 main message abstract updated section 4 with examples question RW: PPSP example: could an object be shared between PPSP and Bittorrent? are you concerned that different swarms could use different schemes -- or different by app -- or both? Börje: both questions Volker Hilt: what are naming scheme requirements are in this context? could we just agreed that we just use strings? why need more in DECADE? Börje: would be good to have only one way to do hashes RW: if you have unique names then you can do client side de-duplication, otherwise have to rely on server-side dedup Volker: if you have to name it differently, then you would have a problem? RW: yes RA: another nice to have property: bittorrent: hash of torrent, concatenated with hash of chunk. Could lead to situations where you overwrite existing content, prevents collisions Ove Strandberg: re PPSP RW: re PPSP vs. DECADE usage -- where does this document live? if requirements are different, do we need to keep different copies of document? does it make sense to have common spec for this? Börje: this is requirements, hopefully we define naming scheme to be used for the latter, want to have only one in IETF we submmited one draft today, can be discussed on ML RW: does PPSP talk about adopting this a WG draft? Yufeng Zhang: urnbis WG could be of interest RA: is this something that we have to wait for? or come with a DECADE solution quickly? Dirk: would need applicability statement for DECADE Dave Harrington: re using something else vs. blocking have concern that IESG might block us if we do not consider existing work if there is no standard, then you can develop one RW: any other questions? RW: how long can the secure names be? Dave Harrington: sounds like security area should be involved Börje: not proposing to do the work here Aaron Falk: where? Börje: not yet decided RW: urnbis maybe Aaron: might be a good idea to get a slot in saag meeting Dave Harrington: also other groups may be doing something similar RW: DECADE not a home for this draft, but has a value here. ---------------------------------------------------------------------- [Audio offset 123:00] Lucy (from Huawei): Integration examples of in-network storage and P2P apps Lucy presenting RW: any questions? RW: how many people have read: a few RW: object to become WG item? no RW: will take to mailing list, but looks good ---------------------------------------------------------------------- [Audio offset: 129:30] Re-chartering discussion RW presenting -- is it time to define DECADE standard? survey document has completed WGLC requirements and architecture documents about to start WGLC three candiate IETF protocols for data access: HTTP, WebDAV, and NFS one candidate IETF protocol for delegating user permissions: OAuth what is requiremd to be a compliant DECADE client? to be a compliant DECADE server? RW: any questions? Haibin: would like to hear opinions David Harrington: re-chartering proposal has to go before IESG need good gap analysis, what works with WebDAV, NFS etc. Would be beneficial to have one MANDATORY to implement. Don't forget deployment environment. Other than that: slide looks really good Aaron Falk: not familar with this work, but like the way it is going. Looks like an application, so wrong area. In apps area there would be other comments? RW: at last IETF, we got moved from APP to transport RA: I know what is deployed in P2P environment. Could you, RW, comment on provider environment RW: did you offer any opinion? RA: from client perspective, would rather not use NFS RW: not sure. David McDysan: since we have a broad range of apps, it would be premature to select one mandatory one now. Could look into detailed use cases requirements after re-chartering? RW: if we do need to take it down to next level of detail, we should do it right now. Investigation should have been done. If we not meet the target then we should go back, not re-recharter David McDyson: some requirements are going much beyond best-effort -- they are quite specific. hopefully we can use same architecture, but may need profiles, different solutions for different environments Akbar: re OAuth, I agree it's a possibility. But have we got to the stage where we concluded that? RW: wanted to find out whether we concluded that. In Survey document, OAuth is mentioned. Would be good to leverage something that exist. Akbar: other alternative that could be considered would be some extension of NFS, HTTP, and WebDAV. RA: re OAuth, OAuth works with server-granted tokens. If we need another way, we may need another model -- Re-chartering discussion (2) Should we make limited set of protocols mandatory? Börje: re only one data transport: support for DTN rules that out RW: yes, we did cover this before Dave Harrington: mission of IETF is to run Internet better with interoperable protocols. We don't prohibit use of anything. If there are different environments, you should be able to use different things. In general, like to see one mandatory one to implement. Aaron Falk: another aspect is that a single protocol that fulfills many requirements may be the same or better than several different protocols -- should we depend on a single name format? -- re-rechartering (4) new topics are essential for the WG? discovery? RA: on discovery: a lot of requirements a single, similar to ALTO, even including 3rd-party discovery. Perhaps we could re-use that. Börje: discovery is important