Thanks Brian, I¹ve been reorganizing the CardSync page [1] and I noticed that the auth part [3] of this section [2] has not yet been updated per your email below. At the least we should mark [3] as being obsolete.
--Paul [1] http://wiki.eclipse.org/CardSync_API [2] http://wiki.eclipse.org/CardSync_API#RESTful_API [3] http://wiki.eclipse.org/CardSync_Authentication On 4/6/09 2:44 PM, "brian walker" <[email protected]> wrote: > Hi All - just to follow up on this topic, we had a breakout discussion on Wed > 1-April-09 to discuss the card synch authentication model. > > Below are the summary points from this discussion: > > * Attendees: John Bradley, Andy Hodgkinson, Valery Kokhan, Paul Trevithick, > Alexander Yuhimenko, Brian Walker > * The following were the proposed decision points from this discussion. Please > reply if there are additional thoughts or ideas related to the below: >> * The authentication model will be based on the model used for Amazon header >> cononicalization ( >> http://docs.amazonwebservices.com/AmazonS3/latest/index.html?RESTAuthenticati >> on.html). We would use "Authorization" header instead of "access_token" for >> passing session ID (access token id), but with our own schema name. For >> example we could use HWS (or propose another name) instead of BASE, DIGITAL >> or AWS. >> * The initial authentication request can be RST based via default p-card >> presented to the back-end card servers as part of the synchronization >> process. To support this we would need to develop the following: >>> * Add a call for getting CardSynch security policy, for example >>> GET/cardsynch/SecurityPolicy. >>> * Define SamlTokenCredentialTO (or another name if a better one is >>> appropriate) and POST it to /cardsynch/rs/AuthCredential. >> * We would use session ID instead of SAML token for per transaction >> authentication. Per >> http://docs.amazonwebservices.com/AmazonS3/latest/S3_Authentication.html, AWS >> does not support session key pair. We could continue to use session ID >> however (aka access token ID) in CardSynch, but would need to add maxLiveTime >> to AccessToken TO. The main reason to use Session ID instead of SAML token is >> due to comparative large size of SAML token. Authorization header may look >> like: "Authorization: HWS session Id". > * There is a follow up action item for Alexander to ensure that this > methodology is also adopted in the serialized selector that is currently being > designed. Stay tuned for a separate email on that topic later this week. > > Please reply if there are additional comments or suggestions. > > regards...Brian > > On Mar 27, 2009, at 3:39 PM, Brian Walker wrote: > >> Ok - we'll aim for 11am EDT on Monday the 30th. Skype will be preferred >> channel - but if someone on below list does not have Skype just let me know >> and I can send out a call in number. >> >> >> Brian Walker >> =brian.walker >> VP of Engineering >> Parity Communications Inc >> cell: 781-801-0254 >> [email protected] >> >> >> >> >> On Mar 26, 2009, at 4:34 PM, John Bradley wrote: >> >>> 11am EDT is fine with me. >>> >>> John B. >>> On 26-Mar-09, at 1:22 PM, Brian Walker wrote: >>> >>>> Ok - we could try for 11am ET if that works for the folks referenced below? >>>> >>>> >>>> Brian Walker >>>> =brian.walker >>>> VP of Engineering >>>> Parity Communications Inc >>>> cell: 781-801-0254 >>>> [email protected] >>>> >>>> >>>> >>>> >>>> On Mar 26, 2009, at 2:24 PM, John Bradley wrote: >>>> >>>>> That overlaps with the OSIS call I wouldn't be able to make it. I am OK >>>>> with before the schemas call. >>>>> >>>>> John B. >>>>> On 26-Mar-09, at 10:05 AM, Brian Walker wrote: >>>>> >>>>>> To help close on this good and relevent topic as it applies to our >>>>>> Selector Harmonization design and implementation - I would propose to >>>>>> host a call (skype call) on Monday 30-March at Noon ET. >>>>>> >>>>>> Anyone/everyone is welcome to join, but at a minimum, would like to have >>>>>> on this call, Paul, Axel, Alexander, John Bradley, Jeesmon, and Andy >>>>>> Hodgkins (if available). Are people from this group available Monday >>>>>> 30-March at Noon ET? >>>>>> >>>>>> thanks...Brian >>>>>> >>>>>> >>>>>> Brian Walker >>>>>> =brian.walker >>>>>> VP of Engineering >>>>>> Parity Communications Inc >>>>>> cell: 781-801-0254 >>>>>> [email protected] >>>>>> >>>>>> >>>>>> >>>>>> On Mar 26, 2009, at 12:35 PM, Alexander Yuhimenko wrote: >>>>>> >>>>>>> Hello Axel, >>>>>>> >>>>>>> CardSync didn't invent something new :) >>>>>>> It's just going to expose high level API via JAX-RS or JAX-WS web >>>>>>> services or XDI endpoint, or ... for synchronizing user data (icard, >>>>>>> history, categories, ...) between few installed user selectors by using >>>>>>> well known design patterns and techologies. I'd like to design it for >>>>>>> supporting data binding to different formats XML, google protobuf, X3, >>>>>>> .... >>>>>>> >>>>>>> Adding authentication by IdP token is tiny fix. >>>>>>> But I thing we have to support few authentication ways. For example, If >>>>>>> user install new selector how/where this selector get ICard for >>>>>>> requesting token? >>>>>>> >>>>>>> I'd like to use single signon/logout auth work-flow by few reasons, >>>>>>> one of them is performance. >>>>>>> Depends on selector synchronization strategy, selector may need update >>>>>>> CardSync store immediately after every change on client side. For >>>>>>> example if client login on RP with pcard (selector generate IdP token >>>>>>> itself), selector must update card history in CardSync store, but for >>>>>>> authentication on CardSync server it should generate second IdP token. >>>>>>> Generating pcard token on client side may take 1-2 seconds, but if your >>>>>>> client work on iphone or android, it will be not so fast. >>>>>>> >>>>>>> Of course we would re-use saml token instead of CardSync access token >>>>>>> id, but if client make few calls for updating CardSync store (for >>>>>>> example updating card categories), it must send saml token every time, >>>>>>> but size of saml token in hundred times more than accesstoken id >>>>>>> (just few bytes). >>>>>>> >>>>>>> The second reason is security. We may restrict number of active user >>>>>>> session. >>>>>>> >>>>>>> >>>>>>> >>>>>>> I'm sorry, we had to share not complete draft version of CardSync :( >>>>>>> But some suggestions were interesting and helpful. if you would >>>>>>> recommend related specifications and/or solutions it'll be useful. >>>>>>> >>>>>>> -- >>>>>>> thanks, >>>>>>> Alexander Yuhimenko <[email protected]> >>>>>>> >>>>>>> On Thu, 26 Mar 2009 16:28:19 +0100 >>>>>>> <[email protected]> wrote: >>>>>>> I think it is more work to make it right later and it is sending the wrong message to other developers if Higgins is not eating it's own mousefood. I have no problem with username/password if this is implemented faster than X509 and as I said the difference to the current UsernamePasswordCredentialTO compared to sending a RST and digesting the saml-token is small Another question: Does the current Higgins architecture allow the selector to access multiple cardstore in parallel? I implemented that openinfocard can have the cardstore on a nfc-phone but if I want to have several cardstores in parallel things get more complicate than I want to tackle with today. I want to use the "standard" cardstore but if I put the phone on the nfc reader then I want this card to be selectable in the selector immediately. Next I want to move cards from one store to the other... -Axel ________________________________ Von: [email protected] [mailto:[email protected]] Im Auftrag von John Bradley Gesendet: Donnerstag, 26. März 2009 16:12 An: Higgins (Trust Framework) Project developer discussions Betreff: Re: [higgins-dev] CardSync : Agenda for Higgins developer call March26 atnoon ET +1 I think I have made this point several time re authn to the remote card store. Last night I was pointing out to Drummond that the selector could easily have a X.509 cert/keypair that is used to back the authn card for the store. Axel I think the idea is under consideration for a future design but people want to get something out the door quickly. John Bradley On 26-Mar-09, at 5:42 AM, <[email protected]> wrote: Regarding 2. and [2] from the latest Agenda email and last weeks email regarding CardSync: "GET Cards" is only one example of the many bad decisions from http://wiki.eclipse.org/CardSync_Protocol Most methods/calls described in CardSync_Protocol reinvent the wheel! Another example: AuthCredential - "/AuthCredential" Signon You are posting a usernamePasswordAuthCredentialTO but this is sö 90ties... The selector accessing the cardstore in the cloud is just a rich-client that needs an access token to access the cardstore. So the way this MUST be achieved to adhere to the overall Higgins goals is: - send a security token request to the cardstore idp and retrieve a security token. Authenticate to the IdP using what ever method it accepts but use the standard RST for this. - use that security token in to access the cardstore (RP) to e.g. retrieve the cards The difference between the data exchanged in the Signon example and the standard RST is minimal. And Higgins already has all the code needed to do it the standard way. -Axel -----Ursprüngliche Nachricht----- Von: [email protected] [mailto:[email protected]] Im Auftrag von Alexander Yuhimenko Gesendet: Donnerstag, 19. März 2009 15:28 An: [email protected] Betreff: RE: CardSync was: [higgins-dev] Agenda for Higgins developer callMarch 19 Hello, GET /cardsync/rs/Cards was implemented just for FC2 demo, it's temporary method due to i didn't finish method which return all changes (ICards, Card history, Card categories, ...) from server between client and server root revisions. if you think "GET Cards" is useful, i may re-implement it by using RoamingStore and card extension for transferring CardSync specific data (Revsion information etc) or without it. I'm sure, CardSync should be compatible with "Identity Selector Interoperability Profile" as much as possible, but it will not be clear/easy always usage extensions for transferring CardSync specific data, so CardSync have to define own schema. it may make sense to support few methods for importing cards from Crd/Crds files and getting them, for example some Card selectors would use CardSync like on-line backup service with minimal changes. -- thanks, Alexander Yuhimenko From: "[email protected]" <[email protected]> Date: March 19, 2009 4:31:56 AM EDT To: "[email protected]" <[email protected]> Subject: CardSync was: [higgins-dev] Agenda for Higgins developer call March 19 Reply-To: "Higgins (Trust Framework) Project developer discussions" <[email protected]> Sorry for chiming in late but why do you define your own cardstore xml in http://wiki.eclipse.org/CardSync_Protocol ? Why isn't the response to GET /cardsync/rs/Cards HTTP/1.1 a RoamingStore as defined in identity.xsd ? http://schemas.xmlsoap.org/ws/2005/05/identity/identity.xsd There are enough extension points in the schema if you need extra data and if you do not want to use some required value then like BackgroundColor then set them to 0. -Axel ------------------------------------------------ Von: [email protected] [mailto:[email protected]] Im Auftrag von Paul Trevithick Gesendet: Donnerstag, 19. März 2009 05:19 An: higgins-dev Betreff: [higgins-dev] Agenda for Higgins developer call March 19 Logistics Time: noon Eastern Dial-in: 1-866-362-7064 / 89-2048# Agenda 1. [Brian] 1.1M6 - was targeted for March 18 See [1] 2. [Brian, Alexander, Andy] Selector Architecture Harmonization See [2] Andy didn't quite get the Synchronizing CardStore done; may need to hand this off to another developer 3. [Mary, Jeesmon, Jochen] Current status of RCP Selector use for EclipseCON demo 4. [Markus] IdAS Authentication Markus is working on a proposal [4] 5. [Paul] Higgins website improvements [3] There are 6 items to discuss 2.1 Flatten Solutions into the 3 solutions areas (nav structure change) 2.2 Reorganizing Solutions 2.3 Eliminate "Higgins Data Model" (nav structure change) 2.4 New solution names 2.5 New component set names 2.6 New Downloads Page Updated home page implementing (2.1, 2.2 and 2.3) is available in the staging area [5] 6. [Paul] FC2 Higgins Workshop Mar 11-13 in Paris We demoed the Selector Linux-GTK reading/writing cards using CardSync FC2 member (Orange) has developed a selector for Nokia J2me phone FC2 has also developed a web selector 7. [Paul] Propose merge web ICM and the Web Selector together into a single solution 8. [Paul] CardSync JohnB's authentication idea (use personal card) Should we use it as the one and only CardStore API for the Higgins selector? This was discussed briefly last week in Paris. Perhaps a simple core protocol and then extensions. 9. [Paul, Markus] All Selectors diagram [6] Discuss if diagram is correct WRT process boundaries (as noted in [6]) [1] http://wiki.eclipse.org/Higgins_1.1M6 [2] http://wiki.eclipse.org/Selector_Architecture_Harmonization [3] http://wiki.eclipse.org/Higgins_1.1_Plan#Website [4] http://wiki.eclipse.org/Authentication_Materials [5] http://www.eclipse.org/higgins/ver2/index.php [6] http://wiki.eclipse.org/Selector_Overview#H1.1_All_Selectors _______________________________________________ higgins-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/higgins-dev _______________________________________________ higgins-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/higgins-dev _______________________________________________ higgins-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/higgins-dev >>>>>>> >>>>>>> _______________________________________________ >>>>>>> higgins-dev mailing list >>>>>>> [email protected] >>>>>>> https://dev.eclipse.org/mailman/listinfo/higgins-dev >>>>>> >>>>>> _______________________________________________ >>>>>> higgins-dev mailing list >>>>>> [email protected] >>>>>> https://dev.eclipse.org/mailman/listinfo/higgins-dev >>>>> >>>>> <smime.p7s><ATT00001.c> >>>> >>>> _______________________________________________ >>>> higgins-dev mailing list >>>> [email protected] >>>> https://dev.eclipse.org/mailman/listinfo/higgins-dev >>> >>> <smime.p7s><ATT00001.c> >> >> <ATT00001.c> > >
_______________________________________________ higgins-dev mailing list [email protected] https://dev.eclipse.org/mailman/listinfo/higgins-dev
