[9:36] Zha Ewry: Not sure if I should be horrified, or happy, that it mostly seems to work
[9:37] Pixel Gausman: 5756 seems to be getting a warm reception in #gridnauts...once people change defaults.xml to use their dotted ip addr instead of hostname
[9:37] Zha Ewry: I'm currently looking at two major, and one minor issue
[9:37] Zha Ewry: The major, major, is the ghosting
[9:38] Movies1963 Beck: if we were in China now that dog that just barked might've been our lunch
[9:38] Zha Ewry: We're failing to clear state fully on logout,and I'm looking at that
[9:38] Zha Ewry: The second, less serious, but anoying, is the "banned" message we sometimes trigger
[9:48] Zha Ewry: Its mostly in three spots, I'll enumerate those, and we'll have them on forge
[9:48] Zha Ewry: I'll be in Europe, so, 9 hours off sync from SLT
[9:49] Zha Ewry: The most reliable way to get my attention will be to my gmail account
[9:49] Zha Ewry: (zha. ewry@gmail.com, it's on my profile)
[9:49] Zha Ewry: While I'm off hopefully recharging the mental batteries.. I have a think/write challange for people
[9:50] Zha Ewry: I've been blythly, asserting, we can do basic proof that Component X, is part of Region Y, and can be trusted, suing certificates, in an SSH or PKIish fashoin
[9:51] Zha Ewry: wherre the basic assertion is we can issue Region Domain "D" a cert "CertD" which it can use to provde to other partners, that it is Region Fomain D, and that Box "Box1" can handshake as well
[9:54] Pixel Gausman: give an example of "Component X"?
[9:54] Zha Ewry: "Service provider Lumpy Labs, issues a Cert, with the followinf properties to Grid Hoster MurkyBusinessModelVW Hosting, MBVWH signs requests.."
[9:55] Tao Takashi: yeah, I would like to write down some scenario using XRDS and OAuth to connect services
[9:55] Tao Takashi: given that I have some time :)
[9:55] Zha Ewry: So..I'm goign to argue the overall design pattern is "Domains" are the basis of trust, "Agent Domain, Regoina Domain, Service Domain" for example
[9:55] Zha Ewry: Where you give a domain a Cert to use as the acnhor of the process
[9:56] Zha Ewry: (You have a not for public consumtpion protocol between the Region Domain and the memebers, which lets them get handed temp certs to prove they are part of the Domain
[9:57] Zha Ewry: So.... when you want to find out if software compoennt X, (say an asset store on a regoin) is part of MBMVWH, it says "I am, and gets atag from the domain, which is can use, short term to prove it is)
[9:57] Tao Takashi: would like a model where there actually aren't "big" domains but more individual services which are easy to replace
[9:57] Tao Takashi: like where the development of the web is going
[9:59] Zha Ewry: (You deperartlye don't want to have a constant update stream "Zha's Region Domain is adding Host1287" to the Domain
[9:59] Tao Takashi: yeah, I guess something like this is needed to keep it manageable
[9:59] Tao Takashi: so that in the end you can point to some trusted node which has contracts with all the other services in order to delegate this decision to it
[9:59] Zha Ewry: "Tao's Turtonic Hosting has rmeoved Data Portability Portal 5"
[10:00] Tao Takashi: I am wondering if something like this could be an extension to OAuth
[10:00] Latha Serevi: I always just imagine these protocols would be the domain cryptographically signing a message of a particular sort, "I authorize MBMVW, public key X, with a temporary permission to do X." It seems straightfoward that once we agree on everybody's pubkey-to-identity mapping, we can use short chains of these signed permission-slips to get various stuff done. How does that mental model fit with y'all?
[10:01] Tao Takashi: as you might already face the same problem on the web if you you e.g. want to read data from 10 services during the signup to one new service. you might not want to tell every single service that it's ok separately. Then on another thought this might be a different problem ;-)
[10:01] Zha Ewry: That's the ballpark I'm thking of
[10:01] Tao Takashi: because it's user centric while the domain thing is service centric
[10:01] Zha Ewry: There has been some fairly cogent concern tha PKI isn't fully up to it
[10:01] Zha Ewry: This is 90% aimed at the server side
[10:03] Zha Ewry: We'r enot looking at doing sigbning for most messages, so we're not trying to bite off all of PKI
[10:03] Latha Serevi: So far, we have two bases to start from -- my first-principles PK auth approach, and OAuth. Are there any others I should be aware of?
[10:03] Zha Ewry: We can assume CAPS and TLI for the security once we've established trust
[10:04] Tao Takashi: Goldie: I am not sure it fits but it would be great if it could be made to fit e.g. by some extension because this is where the web is heading and it would make sense to use such protocols if possible
[10:04] Goldie Katsu: I'll take a look. (I'm assuming you're talking OAuth)
[10:07] Zha Ewry: Todays will add some more deubgging
[10:08] Latha Serevi: I think it woudln't hurt to have more than one model of the underlying identity system, and let the participants (domains/sims/users) be able to choose what list of supported methods they'll handle. L$ banking may go fully crypto-signed-only; most will be happy faster-and-looser, say, any SSL connection to someone on my friendly-hosts list is fine. Will need to beware of this flexibility creating security holes, but it seems "in the spirit" of supporting various approaches.
[10:10] Tao Takashi: Latha: to have different ways of authenticating is already thought about in the spec I think
[10:11] Tao Takashi: as for different ways of authorizing services to do things maybe OAuth can really help
[10:11] Latha Serevi: Which is the relevant spec, by the way?
[10:11] Tao Takashi: as you establish a temporary permission on the consumer side basically like flickr or youtube does these days