RE: [ntpwg] [dhcwg] Re: Network Time Protocol (NTP) OptionsforDHCPv6

["Benoit Lourdelet (blourdel)" <[e-mail]>]

 Danny,


I want to highlight that the  UWisc/Netgear happened without DHCP and I
suspect that a similar incident 
 may happen again. DHCP as you described it could be an amplification
factor but with millions of CPE wrongly configured
 there is no need of DHCP to have things going wrong.

Correctly deployed, an off-the-shelf SOHO router should not have any
server hardcoded -You will never prevent people
 from doing bad thing on a CPE router anyway - and should receive its IP
services server (including NTP) from the provider.

For all other cases, Enterprise where the DHCP server is managed by IT
people of the same administrative domain and ISP settop boxes with DHCP
managed with the Telcos, an DHCPv6 NTP hardcoded address is not going to
hurt anybody as a DHCP Client wont use an expired IP service address for
ever.

Mandating the use of site local address may complicate deployments in
the case of company mergers in the initial stage.
Again, the DHCP server is managed by people of the same administrative
domain that are supposed to configure it with 
 correct IP addresses.

The way NTP is deployed in many SP networks is to use a hierarchy of NTP
server, so that the NTP server is very close to 
 NTP Client. In many case the client is pointing to an NTP server in the
same Point Of Presence.

In enterprise deployments, if IT bothers relying on NTP, they will
deploy their own NTP servers and wont rely on public
 NTP servers. Needless to say that top of the hierarchy enterprise NTP
servers wont use DHCP for configuration.



Benoit

> -----Original Message-----
> From: ntpwg-bounces+blourdel=[e-mail] 
> [mailto:ntpwg-bounces+blourdel=[e-mail]] On 
> Behalf Of Danny Mayer
> Sent: Monday, November 26, 2007 5:41 AM
> To: Ted Lemon
> Cc: [e-mail]; [e-mail]; Richard Gayraud (rgayraud)
> Subject: Re: [ntpwg] [dhcwg] Re: Network Time Protocol (NTP) 
> OptionsforDHCPv6
> 
> Ted,
> 
> Let me try and outline the problem again and please come up 
> with an idea which solves this.
> 
> 1) The DHCP environment is divided into essentially two 
> groups: Hardware like Netgear and Linksys routers and 
> Software like ISC's DHCP Server and Nominum's Dynamic 
> Configuration Server. IETF doesn't allow you to create a 
> protocol which differentiate between these cases.
> 
> The software side of the DHCP implementations are usually run 
> by organizations for their internal use and are actively 
> maintained. I have few worries about these since it's easy to 
> deal with (relatively
> speaking) errors that the sysadmins make.
> 
> The SOHO routers are different since the DHCP servers are 
> built into the firmware and shipped in their 10's of 
> thousands to individuals and small businesses who want 
> wireless connections and routers but don't want to be in the 
> business of configuring and maintaining them.
> 
> So let's say Acme Routers ships a router with a builtin DHCP 
> server which provides NTP server addresses to provide to the 
> DHCP clients and they put just one address in it. Now say 
> Starbucks gets all excited about how cheap they are and buys 
> them for all their coffee stores. Now you have DHCP providing 
> and amplication DDOS attack as all of those people sitting 
> there laptops are all set up with the same NTP server address 
> and sending NTP packets to the same NTP server. Note that in 
> the UWisc/Netgear incident it was the NTP server built into 
> the router that was the problem but it was only one server. 
> Here we are having the router distributing the address to 
> other systems which then do the dirty work and you'd get 10 
> times the effect of a Netgear incident. This is the problem 
> that I'm trying to solve or rather mitigate.
> 
> I refer you to the UWisc/Netgear incident paper that Dave 
> Mills and Dave Plonka wrote:
> http://www.eecis.udel.edu/~mills/database/papers/ptti/ptti04a.pdf
> The brief slide version is here:
> http://www.eecis.udel.edu/~mills/database/brief/ptti/ptti04.pdf
> It also discusses the loads on a number of other servers 
> inclusing NIST and USNO
> 
> PHK's incident with D-Link is written up here:
> http://news.bbc.co.uk/2/hi/technology/4906138.stm
> 
> I await your suggestions on how to prevent the routers 
> becoming amplifiers via DHCP to bombarding NTP servers.
> 
> Danny
> _______________________________________________
> ntpwg mailing list
> [e-mail]
> https://lists.ntp.org/mailman/listinfo/ntpwg
> 

_______________________________________________
dhcwg mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/dhcwg