Re: Late comments on the msgtrk documents

Sorry for the delay.  I've taken Steve's suggestion and added the
following paragraph into section 4.2:

           In some cases, the total length of (local-envid + fqhn + 1)
      (for  the  `@'  sign)  may exceed the total acceptable length of
      ENVID (100).  In this case, the fqhn SHOULD be replaced  by  the
      SHA1(fqhn)  encoded  into  BASE64.   After encoding, the 160 bit
      SHA-1 will be a 27 octet string, which limits local-envid to  72
      octets.  Implementors are encouraged to use an algorithm for the
      local-envid that is reasonably unique.  For example,  sequential
      integers have a high probability of intersecting with sequential
      integers generated by a different host, but a SHA-1 of the  cur-
      rent  time  of day concatenated with the host's IP address and a
      random number are unlikely to intersect with the same  algorithm
      generated by a different host.

A bit wordy, but I wanted to try to explain it clearly.

Comments?

eric



============= In Reply To: ===========================================
: From:  Steve Hole <[email protected]>
: Subject:  Re: Late comments on the msgtrk documents
: Date:  Thu, 18 Oct 2001 16:38:58 -0700

: 
: On Mon, 15 Oct 2001 16:57:53 -0700 Eric Allman <[email protected]> 
: wrote:
: 
: > 
: > Interesting idea -- create (locally) the full
: > 
: > 	[email protected]
: > 
: > which can be arbitrarily long, and then SHA1 it so you get something
: > manageable.  This would make it fit into the existing definition of
: > ENVID.  Unfortunately, the failure mode could be nasty, assuming
: > the server discards duplicates.
: 
: I would suggest that we make a recommendation that FQDN is a good 
: candidate for generating uniqueness.   Further, if the FQDN can't be used 
: directly because it is too long, then a SHA1 hash of the FQDN is a 
: suitable substitute.   It is further recommended that the application 
: create the lhs of the envid such that it conveys uniqueness over a large 
: set of possibilities.  The combination is *very* likely to be unique.
: 
: If the fqdn can't be used because it is not reliable or doesn't make sense
: (the allocated address scenario), then it is left to the application to 
: provide unique information. 
: 
: Cheers.
: ---
: Steve Hole
: Chief Technology Officer -  - Billing and Payment Systems
: ACI Worldwide - MessagingDirect
: <mailto:[email protected]>
: Phone: 780-424-4922
: