Was: Re: Naming and site-local addresses

Keith Moore wrote:
> 
> > If so, how do they cope with interface failure tearing down TCP
> > sessions, other than just failing.
> 
> well, it depends on the application.  for instance, SMTP can just
> detect that the connection was broken and retry sending the entire
> message.  but SMTP by its very nature buffers messages and has
> explicit acks, so having SMTP cope with interface failures
> requires little or no extra effort.
> 
> for that matter, most SMTP sessions are fairly brief.  apps that
> use only brief connections won't experience many errors due to
> interface failures.  other sources of failure are often more
> significant.

Again we consider the problem of address selection for long-running apps...


Here are three models for address selection when both site-local and global
addresses are available.  Which (if any) is preferred:

* Leave it up to application, but recommend preferring global addresses:
  Pros:
    Pick the address with global scope.
  Cons:
    If we are using site-locals because global addresses are unreliable,
    then we are picking the less robust address.

* Leave it up to application, but recommend preferring local addresses:
  Pros:
    Site local addresses may persist where global addresses come and go.
  Cons:
    Default selection is not globally scoped.

* Provide addresses to application 'in order', defaulting to global first. 
Networks that consider their global addresses 'transient' may change
priorities to favour site local addresses.
  Pros:
    Doesn't require smart applications, but applications can override if
    they want.
  Cons:
    Need mechanism to propagate address selection preferences to hosts.


Note that short-running apps (say 10 minutes or less, to pull a number out
of the air) don't really care which address they use as long as it is valid,
which means these apps can reasonably prefer globals in multi-homed
situations (as long as the address is still within it's 'preferred'
lifetime.

Though there is the problem of how long to cache an address for.


-- 
Andrew White                [EMAIL PROTECTED]
--------------------------------------------------------------------
IETF IPng Working Group Mailing List
IPng Home Page:                      http://playground.sun.com/ipng
FTP archive:                      ftp://playground.sun.com/pub/ipng
Direct all administrative requests to [EMAIL PROTECTED]
--------------------------------------------------------------------

Reply via email to