[cap-talk] capability networks compared with ACL networks?

[John Carlson]

I just finished reading:

http://conferences.sigcomm.org/hotnets/2005/papers/argyraki.pdf

That was the first time I have heard of a DoC (Denial of Capability)  
attack.
The gist of it is that you send enough capability requests to prevent  
others from
sending capability requests.  I guess the question remains how often  
you need
to send a capability request.  Is this typical something that happens  
a lot in
capability networks?  I would think that the capabilities might be  
stored for long
term use on clients.  What is the practice?

John
On Feb 5, 2009, at 9:01 PM, John Carlson wrote:

>>
>> One thing that bothers me about his essay is that it only addresses  
>> file
>> access. IMHO, file access is a relatively narrow and uninteresting  
>> part of
>> the access control problem. More important is access control for  
>> active
>> entities, call them servers, daemons, databases etc. These include  
>> things
>> such as CVS, MySQL, Apache etc. etc. Perhaps he will discuss them  
>> when he
>> discusses the setuid bit. It still seems likely to me that the  
>> result will
>> be a separate form of access control for active entities, with  
>> different
>> syntax and semantics. Oh well.
>
> Don't forget access to network ports--perhaps a mixture between a  
> file and an active entity.
> In particular, I am thinking of bind, where a port is bound to an  
> active entity.  I only know
> Berkeley and Unix sockets, ideas from other networks would be  
> interesting.
>
> Maybe everything should be thought of as a port--some place where  
> active entities service and get serviced, and
> where active entities rest.
> Or perhaps you prefer socket--where things hook up to get the  
> current flowing.
>
> Are there any ideas from the OSI model which have been forgotten?
>
> John