Date: Thu, 6 Sep 2001 15:34:14 -0700
From: "Christian Huitema" <[EMAIL PROTECTED]>
Message-ID:
<[EMAIL PROTECTED]>
| Let's try to be practical. We
| have a simple problem to solve, the definition of the flow label in the
| IPv6 specification.
Yes, good.
| I propose the following compromise:
Unfortunately ...
| I believe it captures most of our discussions: we get a basic label that
| is roughly equivalent to the label currently assumed by RSVP/Intserv;
Roughly yes, but very roughly. The flow label used to be 28 bits (all
but the version in the first word). It is already down to 20 bits.
Now 4 more bits are to be removed and used as a label type field for label
types that we have no idea will ever be required?
That seems wrong to me. With a 20 bit field, and a PRNG value in the
field, then there's room for a million different flows, all passing through
a router, and all being uniquely matched by just the flow label (with
address verification later). My guess is that a million is already too
small, but it is also most probably not so far away from what's needed
as to be unusable (to use the million we'd need rather more flows than
that, because we know we'll get collisions, which is what saves such a
small number).
Your proposal reduces that to 64K flows. That I know is way too small.
If we revert the flow label field back to being strictly a flow identifier
set by the source host to identify its flows, then we get a nice clean
definition that can be used by anyone and everyone.
That's what I'd like to see happen.
kre
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
