The IP Over NBMA (ION) Archive

Cell Relay Retreat>ION Archive>month:1997-Dec> msg00015



[Date Prev][Date Next][Thread Prev][Thread Next]  
  [Date Index][Thread Index][Author Index][Subject Index]

SCSP MIB: comment on server IDs

  • From: cliff wang <cliff_wang@VNET.IBM.COM>
  • Date: Fri, 05 Dec 1997 19:38:35 -0500
  • Cc: hoylen@citr.com.au
  • Organization: IBM

> These are some comments about <draft-wang-ion-scsp-mib-00.txt>:

> 1. Permitted lengths of server IDs are limited to 64 bytes.  Since the
>    SCSP stores the length of serverIDs in one byte, the length
>    of the server ID can be up to 255 octets.

Well, when I was designing SCSP MIB, I have already come across the
same question as how long the serverIDs should be. The reason to 
limit to 64 octets are follows:
1) According to RFC 1902, Section 3.5, 

  "An OBJECT IDENTIFIER value is an ordered list of non-negative
   numbers.  For the SNMPv2 framework, each number in the list is
   referred to as a sub-identifier, there are at most 128 sub-
   identifiers in a value, and each sub-identifier has a maximum value
   of 2^32-1 (4294967295 decimal)."

   This restrict the serverID length to be smaller than 128 in MIB.
2) From practical stand of point, a long octet string in MIB is
   very wasteful both in memroy assumption, CPU cycle for table
   entry lookup and for SNMP PDU efficiency. We want to use short
   octel strings if possible.
3) The best solution to solve this in-consistency is to limit the
   serverID length to 6 bits in SCSP protocol and I will propose that
   to SCSP authors at coming IETF. As a matter of fact, an address
   length of 64 bytes seems to be long enough.
 

> 2. The server IDs are represented by both a length INTEGER and an
>    OCTET STRING.  The length is redundant, since it is implicit in
>    the OCTET STRING.  Thus, things like:

>      ScspLSEntry ::= SEQUENCE {
>          scspLSIDLength                    INTEGER,
>          scspLSID                          OCTET STRING(SIZE(0..64)),
>          ...
>    should be just
>      ScspLSEntry ::= SEQUENCE {
>          scspLSID                          OCTET STRING(SIZE(0..255)),
>          ...

Agree, Ken White of IBM has also pointed it out to me earlier. I will
take it out in the next revision.

> 3. Perhaps this should be expressed as a textual conversion so that   
>    all usages of server IDs are consistent.

> e.g.
>        ScspServerID ::= TEXTUAL-CONVENTION
>                        DISPLAY-HINT "1x"
>                        STATUS  current
>                        DESCRIPTION
>                            "The SCSP identifier used by the network entity."
>                       SYNTAX   OCTET STRING (SIZE(0..255))
>                                  -- 
Yes, it can be defined as a textual conversion.

*************************************************
  Cliff Wang                                    
  IPSEC/IPV6 team,  Dept.L04A/NHD               
  International Business Machines Corp.         
  800 Park Offices Drive, G418/B664             
  Research Triangle Park, NC 27709              
  Email: cliff_wang@vnet.ibm.com                
  Voice: (919)486-1255 (IBM T/L 526-1255)       
  Fax:   919-254-0391                           
*************************************************