>
>At Wed, 01 Jul 2009 12:16:12 +0800,
>Song Haibin wrote:
>> >I actually think this is a bad idea. Previous experiences 
>with trying 
>> >to design access control systems (see WebDAV ACLs) have 
>turned out to 
>> >be really overcomplicated. As far as I can tell, this isn't 
>necessary 
>> >for the SIP use cases.
>> >
>> 
>> As RELOAD is designed to support various usages. Other 
>usages may have 
>> this requirement. For example, we can image group communication 
>> scenarios, only members of a group can retrieve and modify the files 
>> belonging to this group. I think ACL might work.
>
>Sure. I'm not saying that one couldn't do an ACL extension to 
>RELOAD at some point in the future to support some future 
>usages. My point is that we don't need it for the current set 
>of applications.
>
>
>> Could you tell me what causes the WebDAV ACLs overcomplicated?
>
>That people want to be able to specify complicated policies, 
>so once you open the door to ACLs it gets complicated.
>And of course the situation here is inherently more 
>complicated since you're asking people you basically don't 
>trust to enforce access control for you.
>

Thanks for explaining. I agree. I guess a shared group key may be a better
choice, at least in p2p scenarios, it does not depend on a third party that
you don't trust to enforce the access control.

>
>> >> Req. 12: It SHOULD be possible to limit the impact of badly 
>> >> behaving P2PSIP nodes on the overall system security.  
>There SHOULD 
>> >> be an option to identify malfunctioning or badly behaving nodes,
>> >and exclude
>> >> or reject them from the P2PSIP system.
>> >
>> >Hmm... It seems to me that this is already possible at 
>least in some 
>> >sense: you use short-lived certificates and then refuse to reissue 
>> >their certificates. Did you have something else in mind? CRLs?
>> >
>> I'm not sure about the solution here. I think it may be hard to 
>> determine the TTL for such kind shor-lived certificates. Misbehaving 
>> nodes still function in the overlay until its certificate 
>expires. Too 
>> short TTL may overload the CA anyway. I don't know how CRLs work in 
>> the overlay, in a p2p fashion or in a c/s fashion.
>
>Me neither. And since overlays are resistant to a certain 
>amount of misbehavior, I don't know if we need to solve this right now.
>

I prefer to mention the consideration to this problem in the security
considerations of the base draft.


Haibin


>-Ekr
>

_______________________________________________
P2PSIP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/p2psip

Reply via email to