Hi,
I fully agree with Itojun's draft on generating the flow label within the
network code. But
I see no reason, why it should not be (at least) readable from application
layer and why
a single UDP or TCP session should not handle more than one streams.
Regards,
Jochen
Hi.
I absolurely agree that it is very useful for the network to be informed that
all the packets of a particular stream belong to a flow that should be treated
similarly (like going along the same path). Clearly the effects of the proposals in
your draft and Itojun's draft at the network level are similar but the network code
can do a much better job of generating the pseudo-random numbers needed to identify
the streams (otherwise you need coordination between applications) and it reduces the
temptation to use the flow label for application layer stuff which is the case when
your proposal is followed.
Regards,
Elwyn
-----Original Message-----
From: Metzler Jochen [SMTP:[EMAIL PROTECTED]]
Sent: Wednesday, November 29, 2000 2:39 PM
To: Davies, Elwyn [HAL02:HG00:EXCH]
Cc: '[EMAIL PROTECTED]'
Subject: WG: Usage of IPv6 flow label
1. draft-metzler-ipv6-flowlabel-00 - the mechanisms proposed in this draft
are not something that we should be supporting - it proposes polluting a network layer
header with application layer info. If an application needs to multiplex streams onto
a single socket in the way proposed it could equally well format the PDU in an
appropriate way without layer pollution.
you missed the thing that it is not the intention of that draft to pollute the
network layer with application layer information. you are right in a way
that an application can / should handle multiplexing of streams internally.
but
imagine that applications that request a special handling by the network
for every single stream, especially wanting all packets belonging to one
single stream to take the same way through the network. in that case
the application must signal the network which packets belong to which
flow. the flow label would be the right place for doing this, there would
be no mixture of network layers.
as shown by some other protocols (e.g. HC) it is legitimate to use lower
layer information on upper layer protocols. if it is e-t-e or mutable a fully
standardized flow label would help in both cases application and network
design.
best regards
jochen
--------------------------------------------------------------------
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]
--------------------------------------------------------------------
