50 likes | 60 Views
Open Issue: Naming. Currently, defines protocol-specific identities, rather than generic ones Some protocols (SIP, XMPP, …) naturally have user identities in URIs sip:alice@example.com ftp:hgs:pw@ftp.example.com May not be the same as ID used to authenticate (e.g., HTTP/SIP Digest identity)
E N D
Open Issue: Naming • Currently, defines protocol-specific identities, rather than generic ones • Some protocols (SIP, XMPP, …) naturally have user identities in URIs • sip:alice@example.com • ftp:hgs:pw@ftp.example.com • May not be the same as ID used to authenticate (e.g., HTTP/SIP Digest identity) • one authentication identity may allow for multiple protocol (e.g., SIP) identities • No obvious way to encode for HTTP • needs: host, realm (?), username • fake: http:alice@www.example.com • new URI scheme • all ugly • Probably need to define this for each using protocol
Open Issue: Domains and Individuals • Currently, only identities represented as URIs • XCAP has notion of domain match, e.g., sip:example.com • Somewhat complicates matching, but otherwise no architectural implications • Not all authentication systems have notion of user@domain, but many do in practice
Open Issue: Exceptions • Exceptions only useful if domains (or groups) allowed • domains: probably most useful for mid-size organizations • small organizations can enumerate • really large organizations unlikely (give access to all my 753,000 fellow postal employees?) • Two kinds of exceptions: • atomic = always part of the same rule • in particular, must be very careful about replacing just exception element temporarily unsafe • override = different rule, overrides more general rule • Override breaks additive model • privacy-unsafe if rule cannot be accessed
Exceptions, cont’d. • Atomic exceptions simply limit matching and are probably ok • Need to be sure we understand cases like: • Rule 1: example.com except alice@except.com • Rule 2: example.com except bob@example.com • What permissions does Bob get? • If ‘Rule 1’, easy.
Open Issue: Relationship to XCAP • XCAP transport (HTTP) + data (list) + manipulation • GEOPRIV has more general model: • carried in LO may not be HTTP • no requirement that RM-LS interface is HTTP • Tying data and list format together seems unnecessary