Re: [shim6] Problems with ULID (IP) over longer sessions.

[Javier Ubillos <[e-mail]>]

On Thu, 2010-04-29 at 16:20 +0200, Alberto García wrote:
> Hi,
> 
> |  -----Mensaje original-----
> |  De: [e-mail] [mailto:[e-mail]] En nombre de
> |  Javier Ubillos
> |  Enviado el: miércoles, 28 de abril de 2010 14:21
> |  Para: [e-mail]
> |  CC: Zhongxing Ming
> |  Asunto: [shim6] Problems with ULID (IP) over longer sessions.
> |  
> |  Hi folks.
> |  
> |  Some time ago I heard a discussion about a potential issue with shim6.
> |  That the Upper Layer Identifier (ULID), chosen as the first pair of
> |  locators used in the communication, could cause
> |  problems/confusion/somethingelse when those locators where no longer
> |  used by the hosts. I.e not in the locator lists.
> 
> That's not much precise...

Sorry, I was being deliberately vague.
I'm not sure what the details are, and I did not want to rule some
possibility out due to my own misunderstandings. 

> 
> |  
> |  I'm unsure about the details of what would cause the problem or what the
> |  consequences could be.
> |  
> |  My own first reaction is that if a software believes it's communicating
> |  with an IP (a shim6 ULID), it might try to spawn more sockets/flows to
> |  that IP.
> 
> This is completely reasonable. If a Shim6 context exists for a
> communication, Shim6 will use the locators in use for the existing context,
> even if the ULIDs are no longer locators. 
> RFC 5533 even supports establishing a new session for a ULID which is not a
> valid locator, by using other addresses and including the ULID as an option.
> 
> Don't see here any problem. 

Then I'm confused about how the context is used and re-used.
I thought that each new call to socket() would create a new context,
independently of source/destination address pairs.

> 
> |  
> |  Have this kind of issues been discussed previously on this list? ( I
> |  couldn't find any discussions about it).
> |  Have any one on this list some more detailed thoughts/experience about
> |  what could cause issues?
> |  
> |  I'm asking this because a couple of colleges and I are looking at using
> |  alternative ULIDs which hopefully would be more of a match with session
> |  identifiers/FQDNs or similar.
> 
> Well, security issues for these alternative ULIDs should be carefully
> considered.

Ofcourse :)


Thanks for your reply
// Javier Ubillos