P B <[EMAIL PROTECTED]> wrote: > Since FreeRadius looks at the four parameters mentioned above while, > as you mentioned in your other mail the Merit server uses a > different approach I understand that the policy that a server > employs for duplicate requests is different!
In order to properly implement the RFC's, the server should follow what the RFC's say. There isn't much choice there, which makes me wonder about what the Merit authors were reading. > What would be the best way for a NAS to use the > Identifier so that it works with 'most' Radius > servers? This NAS has to also interwork with some > older Radius installations and I am not sure how they > handle retransmission! You follow the RFC's. > Is it OK to : > 1. use the identifier sequentially (0-255 and wrap) Yes. > 2. for every identifier record the timestamp at which > its last response came back. That's not necessary. > 3. while sending sending a request only use > identifiers > which are atleast 'x' seconds old. 'x' can be the > cleanup time of the Radius server! > wont this guarantee that no request with the same > identifier (other than real retransmissions) is sent > to the server within any 'x' second timeframe? If your NAS is busy, it may be sending 1000's of requests a second. So keeping track of timestamps won't help. For any set of 256 ID's, they will ALL have the same timestamp. > I guess the tradeoff is a capacity hit, but is there > any other standard method? The RFC only mentions > using a different identifier in each new request. You can also use a different source port. Alan DeKok. - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html
