This behavior should probably  be in the document.

> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf
> Of Alan DeKok
> Sent: Sunday, February 03, 2013 9:02 AM
> To: [email protected]
> Cc: [email protected]
> Subject: Re: [abfab] [radext] New Version Notification for draft-perez-
> radext-radius-fragmentation-04.txt
> 
> [email protected] wrote:
> > In order to maintain deal with RADIUS clients and servers that do not
> > support fragmentation, it would be cleaner to use RADIUS capability
> > negotiations (draft-halwasia-radext-capability-negotiation) to
> > negotiate use of More-Data-Pending attribute.  This would allow all
> > RADIUS messages to use More-Data-Pending attribute for fragmentation
> > with no exception.
> 
>   The existence of a More-Data-Pending in an Access-Request indicates that
> the client supports fragmentation.  Servers which do not understand that
> attribute will return an Access-Accept / Reject instead of an Access-
> Challenge.  The client will realize that it has not sent all of the data
it wants to
> send, and will terminate the users session.
> 
>   A server supporting fragmentation will start off by sending an
Access-Accept
> with Service-Type = Additional-Authorization.  Clients not supporting
> fragmentation will not recognize the Service-Type, and under RFC 2865
> Section 5.6, treat it as Access-Reject.
> 
>   There is no need for capability negotiation here.  If a client needs to
send a
> lot of data, it either sends all of the data, or terminates the users
session.  If a
> server needs to send a lot of data, it either sends all of the data, or
sends a
> Service-Type which causes the client to terminate the users session.
> 
>   It would be *nice* to know this prior to terminating the session.
> Simply rejecting the user unexpectedly is safe, but unfriendly.
> 
>   Any capability negotiation here is a user interface problem.  It helps
the user
> or administrator understand why a session was terminated.  It doesn't help
> the RADIUS client or server.
> 
>   Alan DeKok.
> _______________________________________________
> abfab mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/abfab

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

Reply via email to