jon@ifi.uio.no (Jon Oelnes) (10/11/90)
In the ISO connection-less network protocol (ISO8473) one function is Route Recording. I have only an old copy (DIS version) at hand, and my question is if the following text is still valid, and if my interpretation is correct: "When Complete Route Recording is selected, PDU reassembly at _intermediate systems_ is performed only when the Derived PDUs that are reassembled all took the same route; otherwise the PDU is discarded, .." Now, reassembly is normally performed in the final _end system_, not in the ISes. My interpretation of the text is that no restriction is posed on reassembly in the _end systems_. An ES may perfectly well reassemble fragments which have followed different routes. Also, the point of discarding (in an IS) fragments which followed different routes, is only valid if the IS really does reassembly. If it just passes the fragments on, no error occures. Comments, anyone - by E-mail if this is not considered to be of general interest. Thanks in advance. Jon Oelnes, Norwegian Computing Center, Oslo E-mail: Jon.Olnes@nr.no or jon@ifi.uio.no
bhk@merak.la.locus.com (Brad Kemp) (10/11/90)
In article <CMM.0.88.655642509.jon@gode.ifi.uio.no> jon@ifi.uio.no (Jon Oelnes) writes: >In the ISO connection-less network protocol (ISO8473) one function is >Route Recording. I have only an old copy (DIS version) at hand, and my >question is if the following text is still valid, and if my interpretation >is correct: [ valid text Deleted ] >Now, reassembly is normally performed in the final _end system_, not in the >ISes. My interpretation of the text is that no restriction is posed on >reassembly in the _end systems_. An ES may perfectly well reassemble >fragments which have followed different routes. >Also, the point of discarding (in an IS) fragments which followed different >routes, is only valid if the IS really does reassembly. If it just passes >the fragments on, no error occures. Unfortunatly fragements which follow different routes and which have the Compelete Route Recording selected will never reach an ES. Lost fragements will also cause the PDU to never reach an ES. It is mandatory for IS's to be able to reaseemble. Although it is never mentioned in the standard, IS's should foward all PDU fragments through the same route if Complete Route Recording or Security is set. Both these functions require the PDU to be dropped if the PDU cannot be reassembled. If the users does not want to lose PDU's due to the fragements travelling different paths the Partial Route Recording should be used. Complete Route Record gives the user the actual path that the PDU traversed. It is interesting to note that if there is not enough room in the head to add the Network Entity Title, conformant systems MAY still attempt to reassemble the PDU and drop it if it cannot be reassembled. However if it can reassemble the PDU, the PDU is forwarded without the addition of the NET. > >Thanks in advance. >Jon Oelnes, Norwegian Computing Center, Oslo >E-mail: Jon.Olnes@nr.no or jon@ifi.uio.no Brad Kemp Locus Computing Corp bhk@locus.com uunet!lcc!bhk
glenn@SLOTH.NCSL.NIST.GOV (K. Robert Glenn) (10/12/90)
>Unfortunatly fragements which follow different routes and which have >the Compelete Route Recording selected will never reach an ES. Why? I don't see the logic here. >Lost fragements will also cause the PDU to never reach an ES. The complete PDU will not reach the ES, but parts of it might causing the ES to generate and send an ER PDU (if permitted). >It is mandatory for IS's to be able to reaseemble. Table 10 in the IS version states that the reassemble PDU function is an implementation option for ISs(Fowarding systems). >Although it is never mentioned in the standard, IS's should foward >all PDU fragments through the same route if Complete Route Recording or >Security is set. Both these functions require >the PDU to be dropped if the PDU cannot be reassembled. >If the users does not want to lose PDU's due to the fragements >travelling different paths the Partial Route Recording should >be used. >Complete Route Record gives the user the actual path that the PDU >traversed. It is interesting to note that if there is not enough >room in the head to add the Network Entity Title, conformant >systems MAY still attempt to reassemble the PDU and drop >it if it cannot be reassembled. >However if it can reassemble the PDU, the PDU is forwarded without >the addition of the NET. > >Brad Kemp >Locus Computing Corp >bhk@locus.com uunet!lcc!bhk > Very little is mentioned in the standard about Security. The only thing I can find concerning the Security function discarding PDUs, is that if an 8473 implementation receives a PDU with the Security option in use and the implementation does not support Security, then the PDU is discarded. K. Robert Glenn N.I.S.T. glenn@osi3.ncsl.nist.gov
jmh@ns.network.com (Joel Halpern) (10/12/90)
In article <18116@oolong.la.locus.com> bhk@merak.la.locus.com (Brad Kemp) writes: >In article <CMM.0.88.655642509.jon@gode.ifi.uio.no> jon@ifi.uio.no (Jon Oelnes) writes: >>In the ISO connection-less network protocol (ISO8473) one function is >>Route Recording. I have only an old copy (DIS version) at hand, and my >>question is if the following text is still valid, and if my interpretation >>is correct: > > [ valid text Deleted ] > >>Now, reassembly is normally performed in the final _end system_, not in the >>ISes. My interpretation of the text is that no restriction is posed on >>reassembly in the _end systems_. An ES may perfectly well reassemble >>fragments which have followed different routes. >>Also, the point of discarding (in an IS) fragments which followed different >>routes, is only valid if the IS really does reassembly. If it just passes >>the fragments on, no error occures. > >Unfortunatly fragements which follow different routes and which have >the Compelete Route Recording selected will never reach an ES. >Lost fragements will also cause the PDU to never reach an ES. >It is mandatory for IS's to be able to reaseemble. >Although it is never mentioned in the standard, IS's should foward >all PDU fragments through the same route if Complete Route Recording or >Security is set. Both these functions require >the PDU to be dropped if the PDU cannot be reassembled. These comments appear to imply some requirements on routers which I was unaware of, and some of which I am not sure I understand. The reference to security requiring that all packets follow the same path does not match the goals of such proposed work as SP3, where between two security routers, packets may follow whatever route they find. In addition, re-assembly at the end router is supported by the fact that the router is intercepting these packets for local processing anyway (under the most general SP3 encapsulation in fact the packets are addressed to the remote security router). I am therefore not clear on what is meant by the reference above to security requiring intermediate re-assembly and common paths. With regard to complete route recording. If the routers must perform re-assembly, is it legitimate for the routers to decide instead to always drop any fragmented packet which is requesting complete route recording? Re-assembling packets on the fly which are not intended for local processing can be a real headache. Thanks, Joel M. Halpern jmh@nsco.network.com Network Systems Corporation P.S. Does anyone know what the current ISPs/GOSIPs say about support for the compelte route recording option?