Dear Mr. Stephane Bortzmeyer,
Thank you for your comments.
The address space of IPv6 is huge means in the future, the next generation
Internet will have a huge number of the IP addresses as well as Domain names.
The time and space complexity claim of section 3 is as follwoing:
Complexity ( similar as the paper in Huang, L., "VIRGO: Virtual Hierarchical
Overlay
Network for Scalable Grid Computing ",Proc.
European Grid Conference(EGC2005), in LNCS 3470,
p911-921, 2005.)
Because the DNS server nodes are virtually organized as a tuple virtual
tree, every DNS server has a route table which includes prerequiste DNS
servers' IP addresses for Tree Paths (TREE portion) and cached DNS servers' IP
addresses (CACHED portion).
Because the message is routed according to the minimum of theoretical
distance from destination node , and the route table contains TREE portion,
every hop reduces the distance from destination node by at least one hop,
Therefore,
hops(a,b) and message_cost < length(a) + length(b)-1 (1)
Where, hops(a,b) is for the hops from node a to node b; length(a),length(b)
are for
node a domain name lengths and node b domain name lengths respectively. for
example, the lenth of www.nic.fr is 3.
So,
time complexity = O(L) (2) ,
where L is the tree depth.
Because the route table of the virtual gateway nodes virtually existed from
root layer to bottom layer groups has the maximum route items of nodes's
information, we have:
items of route table = L*N_tuple*nvg +Max_Cached (3)
,where L is the depth of virtual Tree, N_tuple is multiplicity of gateway
nodes for virtual tree,
nvg is number of virtual groups, Max_Cached is the maximum number of cached
records in the route table
Therefore,
space complexity = O(L) (4)
Suppose all leaf virtual groups have the same number of nodes, nvg and all
non-leaf virtual groups have the same number of nodes, nvg *n_tuple , then the
number of layers L is log_nvg(N) , where N is the total node size of the
network. So,
time complexity and message_cost = O(log N) (5)
space complexity = O(log N) (6)
Why use Cached route nodes in route table?
This is because that tree like network will has traffic in the nodes within
root layer. This can be avoided by randomly choose a next hop node from the
CACHED portion nodes and TREE portion nodes.
2.3 "On-demand domain" section
a machine may "have" several domain names.
Domain names now is used for corresponding with IP address. But, I argue that
in the future, Domain name is not only for IP address, but has some social
meanings in it. For example, a machine is the provider of Song of Britney and
Madanna, this machine may use
www.Britney.popular.music and www.Madonna .popular.music as its Domain
Names. You can easily locate the machine which provides Song of Britney and
Madanna by registering
this machine the domain names of www.Britney.popular.music and www.Madonna
.popular.music, and virtually join the corresponding virtual groups in the DNS
network.
Thank you again.
Best regards,
Lican Huang
huang lican <[EMAIL PROTECTED]> wrote:
From: "huang lican" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Subject: Fw: [DNSOP] Re: draft-licanhuang-dnsop-distributeddns-02
Date: Wed, 17 Oct 2007 10:11:43 +0800
----- Original Message -----
From: "huang lican"
To: "Stephane Bortzmeyer"
Cc:
Sent: Wednesday, October 17, 2007 10:10 AM
Subject: Re: [DNSOP] Re: draft-licanhuang-dnsop-distributeddns-02
> Dear Mr. Stephane Bortzmeyer,
>
> Thank you for your comments.
> The address space of IPv6 is huge means in the future, the next generation
> Internet will have a huge number of the IP addresses as well as Domain names.
> The time and space complexity claim of section 3 is as follwoing:
>
> complexity ( similar as the paper in Huang, L., "VIRGO: Virtual Hierarchical
> Overlay
> Network for Scalable Grid Computing ",Proc.
> European Grid Conference(EGC2005), in LNCS 3470,
> p911-921, 2005.)
>
> Because the DNS server nodes are virtually organized as a tuple virtual tree,
> every DNS server has a route table
> which includes prerequiste DNS servers' IP addresses for Tree Paths (TREE
> portion) and cached DNS servers' IP addresses (CACHED portion).
>
> Because the message is routed according to the minimum of theoretical
> distance from destination node , and the route table contains TREE portion,
> every hop reduces the distance from destination node by at least one hop,
>
> Therefore,
> hops(a,b) and message_cost < length(a) + length(b)-1 (1)
> Where, hops(a,b) is for the hops from node a to node b; length(a),length(b)
> are for
> node a domain name lengths and node b domain name lengths respectively. for
> example, the lenth of www.nic.fr is 3.
>
> So,
> time complexity = O(L) (2) ,
> where L is the tree depth.
>
>
> Because the route table of the virtual gateway nodes virtually existed from
> root layer to bottom layer groups has the maximum
> route items of nodes's information, we have:
>
> items of route table = L*N_tuple*nvg +Max_Cached (3)
>
> ,where L is the depth of virtual Tree, N_tuple is multiplicity of gateway
> nodes for virtual tree,
> nvg is number of virtual groups, Max_Cached is the
> maximum number of cached records in the route table
>
> Therefore,
> space complexity = O(L) (4)
>
> Suppose all leaf virtual groups have the same number of nodes,
> nvg and all non-leaf virtual groups have the same number of nodes, nvg
> *n_tuple , then the number of layers L is log_nvg(N) , where
> N is the total node size of the network. So,
>
> time complexity and message_cost = O(log N) (5)
>
> space complexity = O(log N) (6)
>
>
> Why use Cached route nodes in route table?
>
> This is because that tree like network will has traffic in the nodes within
> root layer. This can be avoided by randomly choose a next hop node from
> the CACHED portion nodes and TREE portion nodes.
>
> 2.3 "On-demand domain" section
> a machine may "have" several domain names.
>
> Domain names now is used for corresponding with IP address. But, I argue that
> in the future, Domain name is not only for IP address, but has some social
> meanings in it. For example, a machine is the provider of Song of Britney and
> Madanna, this machine may use
> www.Britney.popular.music and www.Madonna .popular.music as its
> Domain Names. You can easily locate the machine which provides Song of
> Britney and Madanna by registering
> this machine the domain names of www.Britney.popular.music and www.Madonna
> .popular.music, and virtually
> join the corresponding virtual groups in the DNS network.
>
>
> Thank you again.
>
> Best regards,
>
> Lican Huang
>
>
> ----- Original Message -----
> From: "Stephane Bortzmeyer"
> To: "Lican Huang"
> Cc:
> Sent: Wednesday, October 17, 2007 4:30 AM
> Subject: [DNSOP] Re: draft-licanhuang-dnsop-distributeddns-02
>
>
> > I do not find in the I-D draft-licanhuang-dnsop-distributeddns-02.txt
> > a notice about the right forum for the discussion? Since there is
> > "dnsop" in the title, I copy the dnsop WG but let me know if you
> > disagree.
> >
> > This is not a complete review of the draft because my main problem is
> > that the draft is quite... drafty. There are few technical details,
> > there is a lot of hand waving (saying "IPv6 is huge" each time there
> > is a difficulty will not help), many claims are not backed by facts
> > (such as "the existing DNS architecture is unsuitable for the growth
> > of the Internet" or such as the time and space complexity claim of
> > section 3, which is apparently not demonstrated).
> >
> > Some sections are very puzzling, for instance 2.3 "On-demand domain
> > names" seem to have been written by someone with very little Internet
> > experience because it presents as a big discovery the fact that a
> > machine may "have" several domain names.
> >
> > I find personally very difficult to determine if there is actual
> > substance in this draft. For the next version, I suggest to focus on a
> > few things, and to stop trying to do so many at the same time.
> >
> > Do note, for instance, that the Ramasubramanian and Sirer paper that
> > you mention is careful to talk only about resolution and does not go
> > into registration issues.
> >
> >
> >
> > _______________________________________________
> > DNSOP mailing list
> > [email protected]
> > https://www1.ietf.org/mailman/listinfo/dnsop
---------------------------------
Yahoo! Answers - Get better answers from someone who knows. Tryit now._______________________________________________
DNSOP mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/dnsop
