OGSA Teleconference November 20, 2003 ===================================== * Participants Bill Horn (IBM) Hiro Kishimoto (Fujitsu) Fred Maciel (Hitachi) Takuya Mori (NEC) Andreas Savva (Fujitsu) Frank Siebenlist (ANL) David Snelling (Fujitsu) Ravi Subramaniam (Intel) Jem Treadwell (HP) Jeffrin Von-Reich (HP) Minutes: Andreas Savva * Actions generated in this call - Action: HK to send updated F2F agenda before the F2F. - Action: FS to do - High level analysis & clarification - Incl. grouping or coloring figure to make things clearer. - OGSA doc revision to make clear what's needed for OGSA - OGSA Security revision (WG is currently in holding mode) - Action: FS to send a link to this document/email or to the mailing list archive. >Done. - Action: FS to send out related presentation from the ACM security workshop. >Done. - Action: FS to answer TM's questions, (marked "?" in the security summary document) before F2F. - Action: HK to make sure discussion of the security summary document is on the F2F agenda. * Last week's minutes - approved * Agenda bashing/AOB Agreed to move the Use Case discussion up the agenda to make sure this topic is covered in this call. Note: No teleconference next week (11/27). Next meeting is the F2F at ANL Dec. 3-5. * Use Cases - JvR: Status update - No updates yet! - Waiting for 4 use cases to be re-written in v2 template. - Waiting for new use cases from APPAGG-RG. - Request people to complete revisions by next Friday. - DS: Reality Grid UC delayed because the person doing it was busy with SC03 preparations. Next week Friday may be possible. - AS: Will try for intermediate version (include comments from David Savage maybe inline as issues). Original authors should review so it is unlikely to be the final version. * Interim F2F updates ** F2F #5 (December 2003) - HK sent out the attendees list earlier. - 13 people on the list at the moment - 10 people sent/confirmed details - REMINDERS - If you are coming make sure you send your details or you may not be able to enter Argonne. - Arrange your reservation with Argonne guest house. Local contact name: Ian Foster. - It is NOON to NOON (12/3 to 12/5). - TM: Takashi Kojo would like some time during the interim to talk about CDDLM. - HK: There is a 90 minute "resource management services" section review already scheduled for Thursday afternoon. The CDDLM and provisioning will be discussed there. - Action: HK to send updated F2F agenda before the F2F. - Sec 4.4. (Resource Mgmt) - FM: Thought he owned this section but HK asked Takashi Kojo to write it. There is a first draft in CMM that covers this area so would like to contribute too. ** F2F #6 (Feb 2004) - Date was fixed last time (Feb 9,10) - Exact location hasn't been fixed. - RS: Hasn't been able to find a conference room with internet connection yet. Still talking it over with the IT dept. - HK: Waiting for Andrew Grimshaw to confirm availability of SDSC. The venue for this F2F will probably be chosen during the Dec. F2F meeting. * Security discussion - TM's security summary gridforge url: https://forge.gridforum.org/projects/ogsa-wg/document/Summary_of_OGSA_Security_Roadmap/en/1 - FS: Good summary of the Security documents. Some nits but otherwise it's excellent. - Page by page review (within time limit) - p2: "application specific components" Wrong term. "High level security services" or "security applications" is ok. But not components. - p3, Fig. 2: - The Web Services Standards layer shouldn't appear in this stack. Of course it exists but it is not security specific. - This figure is presented as a stack (and called a stack) but it isn't really a stack. - Two types of stacks: - upper components depend on lower components, or - abstraction stacking - Problem here is that it is not clear which of the two types (if any) it is. Agreed that it is more of a conceptual/abstraction 'stack.' (with reservations on using the term 'stack' since the figure does not really show a layering/dependency relation.) So text at the top of p4: "...high level layers...represents dependency" is wrong and should be deleted. - RS: Not all boxes are the same. There is a mixture of services/protocols; perhaps we should separate them out (to make it clearer) - Some boxes may define both. - Adding color to indicate service/interface/protocol may help people understand this figure better. Action: FS to do - High level analysis & clarification - Incl. grouping or coloring figure to make things clearer. - OGSA doc revision to make clear what's needed for OGSA - OGSA Security revision (WG is currently in holding mode) HK: In Fig. 2 what's the difference between authorization (layer 3) vs authorization service (layer2) - authorization delegation vs asking for authorization - FS: Published model on using XACML in OASIS (outside GGF). The hope is to incorporate in OGSA AuthZ-WG. Action: FS to send a link to this document/email or to the mailing list archive. >Done. http://www-unix.gridforum.org/mail_archive/ogsa-wg/msg00598.html - FS: Trying to come up with a technology to express and implement delegation of rights. There are use cases but not the means to implement them yet. E.g., Authorization assertion to express "it's ok to let 'this entity' access 'this file'" even if the fileserver does not know 'this entity' before hand. - Q: Which is more important delegation or authorization? FS: Authorization call out (Fig. 3) is as important to standardize as expressing delegation rights. - GT3 uses delegation to derive rights to apply to a new account (dynamically created) - Too crude at the moment (all or nothing model). Want to be able to do more refined delegation. Action: FS to send out related presentation from the ACM security workshop. >Done. http://www.mcs.anl.gov/~franks/ACMXMLGridSecurity.pdf - HK: GT3 performance issues - FS: Well aware of it. Currently working with major clients/partners to resolve them. Half of the performance problem seems to be in the XML parsing. Standard tools are not that fast. - Some commercial products may offer faster processing but also make assumptions that may cause interoperability problems. - HK: So is the root of the problem implementation or architecture? - A bit early to tell. Don't want to tinker with the architecture until it is clear what the problem is. (Maybe in 2-3 months it will be clear.) - How about using other bindings? - It is a common story. But the reality is that there is no available tooling right now for any other binding. - Globus Toolkit is enabler. Vendors can add their own value, and perhaps this is one area that they can do so. - It is important to maintain a reference implementation though. - Need care to provide reasonable performance or globus support may become weaker. - Fig.3 is NOT an architecture picture. It is an example of an implementation to show who benefits from standardization of interfaces. - It also shows things that are not services. - Secure conversation link is a collection of technologies not a single thing. HK: Do services used by requestor have to mirror the ones that have to be used by the provider. - Might not necessarily be symmetric but the overall architecture works more along the lines(assumptions) of a peer to peer interaction. - Sometimes it is hard to grasp that authorization callouts are needed on the requestor's site too. It happens now too, e.g., going through a firewall proxy. - p6 fig 4. - Again not a stack. e.g., authorization does not depend on lower stuff. - It's a roadmap figure (out of the IBM/MS docs). Probably don't need this. Consider deleting. - p8 - GSSAPI-SecureConversation - 'overidden' is not correct. WS-SecureConversation (building on WS-Trust) are generic schemes to provide security context. Think of GSSAPI-SecureConversation (written by Argonne team) as one profile. - HK: GSI-WG owns GSSAPI-SecureConversation? - FS: No. GSSAPI and GSSAPI-SecureConversation are different. - GSSAPI mimicked using portypes and exchange of tokens (done as services) partly protocol, also defining token format. - GSSAPI-like not GSSAPI. - p11, Table 5: - Q: High/Medium/Low is TM's evaluation? Yes. - 'high' already implemented in GT3 - Sometimes implementation not necessarily compliant - Would like to have something like this to show clients and describe what is going on. - Action: FS to answer TM's questions, (marked "?" in the security summary document) before F2F. - Action: HK to make sure discussion of the security summary document is on the F2F agenda.