Re: (sipp) comment on draft-ietf-sipp-routing-addr-02

[Ross Callon <[e-mail]>]

> From [e-mail] Tue Aug  2 15:08:17 1994
> Date: Mon, 1 Aug 94 07:11:24 PDT
> X-Sender: [e-mail]
> Mime-Version: 1.0
> Content-Type> : > text/plain> ; > charset="us-ascii"> 
> To: [e-mail]
> From: [e-mail]
> Subject: Re: (sipp) comment on draft-ietf-sipp-routing-addr-02
> Sender: [e-mail]
> Reply-To: [e-mail]
> Content-Length: 594
> 
> Bob and Matt,
> 
> While it's true that 802.x addresses are just one kind, they are likely,
> however, to be somewhat "distinguished".  For example, it is most likely
> that both "serverless" (ad hoc) and "stateless" configuration will be done
> using these addresses (on networks that support 802.x addresses).
> 
> Greg Minshall

But Greg, both serverless and stateless configuration can be done
using *any* pre-assigned "low end part of address" that the host
knows a priori, plus the subnet prefix as the high order part of 
the address. There is no need for the "low end part of the IP6
address" to be bit-for-bit identical with the subnet address.

The **ONLY** reasons (in my opinion) that IEEE 802 addresses make 
sense for use as the low order 48 bits of the IP6 address is to 
avoid having to administer another similar sized space, and to 
avoid having two address proms, or their equivalent, on hosts. 
Also, if the low-end-known-a-priori part of the address is tied 
into the subnet address, then if there is a collision, then this 
will be found and will need to be resolved anyway (in order to make
the link layer work). These are important reasons.

However, people seem to perpetually get confused regarding whether
the use of IEEE mac addresses as an administratively convenient way
to get a high-probability-unique-on-the-subnet identifier means for
example that all systems need to have a physical Ethernet interface,
or ARP can avoid having to do the address lookup, etc. Having run 
into this confusion for years, I think that we should try hard to 
make this non-confusing.

I would therefore be more inclined towards using a single ID space
which is not tied into the subnet address, rather than try to map
every possible subnet address type into an ID space. 

Ross
 ------------------------------------------------------------------------------
IETF SIPP Working Group - Archives:  parcftp.xerox.com:/pub/sipp
Unsubscribe:	unsubscribe sipp		(as message body, not subject)
Direct all administrative requests to [e-mail]