Re: [dhcwg] New Version Notification for draft-jiang-dhc-stateless-reconfiguration-00.txt

[Sheng Jiang <[e-mail]>]

Hi, Yusuke,

Thanks for your use case. We will inlcude the abstract desciption into next version, which will include several suitable use cases.

Also thanks for your opinion on my designing questions. We are still in early state regarding to design the solution. We are open for all kinds of feedback.

Best regards,

Sheng
________________________________________
From: Yusuke DOI [[e-mail]]
Sent: 06 November 2013 5:25
To: Sheng Jiang
Cc: <[e-mail]>
Subject: Re: [dhcwg] New Version Notification for       draft-jiang-dhc-stateless-reconfiguration-00.txt

Dear Sheng, and WG members,

Let me describe the issue I've introduced on the meeting with regard to
draft-jiang-dhc-stateless-reconfiguration-00.

Target system we expect is a wireless mesh network runs over
802.15.4e/g. It contains 1000s of nodes and expects quite low traffic.
There are some configuration parameters (example is MPL parameters --
e.g. draft-doi-roll-mpl-parameter-configuration). At most cases, we
don't expect any update of configuration. However, we have to expect
some instances of mesh networks could go beyond default parameter's
assumption. For example, we need to tweak the configuration if we have a
mesh network with too high/low density, too many e2e hops or have higher
background traffic than expected, etc.

Hence, we need to have a way to update configuration of a network, but
we don't expect frequent change.

Current choice we have in our mind is to use stateful DHCPv6. This
decision is made just because we have no other choice to distribute and
update network-wide configuration. However, stateful DHCPv6 may
introduce following problems.

1) lease renewal traffic
2) update delay

Because a mesh network cannot transmit much traffic, I expect lease time
more than a day or even for a week to reduce the traffic. This makes we
need to wait weeks to propagate configuration update.

Here is my preference to the questions on the draft/ppt.

Question 1: b) seems to be simplest
Question 2: We prefer direct advertisement because mesh networks may not
be able to accomodate traffic from 1000s of nodes to update their
configuration within short period of time.
Question 3: no opinion
Question 4: If just update trigger is sent from the server, I believe
the update window should be specified with a DHCPv6 option. The window
size depends on network configuration and how important the update is.

Best Regards,

--
// Yusuke DOI <[e-mail]>