The MPLS WG Archive[Date Prev][Date Next][Thread Prev][Thread Next] [Date Index][Thread Index][Author Index][Subject Index] What is the bad tspec?
Chetan Pinto wrote: > > Few examples > (1) m should not be 0 No. Zero means "don't care" in all IntServ objects - including TSPEC, FLOWSPEC and ADSPEC. Zero is legal for all fields. > (2) m must not be greater than M > (3) r must not be greater than p These make sense, as long as you keep in mind that M and p can be zero without violating these rules. >> If I set the token bucket rate very large, it will send back the patherr >> msg with 'Requested bandwidth unavailable', it will not send the bad >> tspec. Keep in mind that a huge value in the TSPEC may not cause a PathErr. Until the Resv message arrives from downstream, the actual size of the reservation could be any value between zero and the value of the TSPEC. If the FLOWSPEC in the Resv is too large, then a PathErr is mandatory. The TSPEC in a PathErr message is optional. More specifically, the entire sender descriptor (SENDER_TEMPLATE, TSPEC and ADSPEC) is optional in PathErr. IMO, the sender descriptor should be sent, because you can't uniquely identify a single LSP within a session without it. A PathErr with a SENDER_TEMPLATE, but no TSPEC doesn't seem to be legal, according to my reading of RFC 2205. (It violates the BNF description of a sender descriptor in section 3.1.3.) I wouldn't reject the message, however, since the TSPEC can be determined by looking at the stored value. (Part of being liberal in what you accept.) -- David
|
|