Re: [bcop] Fwd: [Bcop-gc] documentation ipv6 prefix
On 22/08/14 02:56, 马严 wrote:
Hi, Jan and all,
As RFC3849 specified, the prefix reserved for documentation is a /32 block, 2001:DB8::/32 while people can use the following: net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48 we can also use net A = 2001:db8:1::/48 net B = 2001:db8:8000::/48 net C = 2001:db8:a000::/48 for being easy recognized as separated networks.
Yes, I agree, but this is different just to some limited extent. People, not very familiar with IPv6 and on their learning curve might mistakenly understand this as prefixes in one network. To be really sure they distinguish between the networks (being just different local networks or different AS-es) I think completely different IPv6 prefixes should be used, visually different from the first "character" on... I thought this is what Japanese colleagues are suggesting... (Including Seiichi-san to cc:) Cheers, Jan
The only shortcoming that I can think of is, because 2001:db8::/32 is one big block, when being used to describe inter-domain network topology, /32 address block may easily be considered as all networks belong to one organization. Any comment?
I also cc:ed this email to the co-author of RFC3849, G.Huston, Chief Scientist from APNIC, for further discussion.
Best regards, --MA Yan
----- reply email ----- *Sender:*Jan Zorz @ go6.si <jan@go6.si> *Recipient:*bcop <bcop@ripe.net> *Time:*08/21/2014 22:11:55 *Subject:*[bcop] Fwd: [Bcop-gc] documentation ipv6 prefix
Dear RIPE BCOP community,
I got a question from Seiichi Kawamura, JANOG BCOP co-chair and I think this suggestion/question would be best if discussed here on this mailing list (and maybe also on IPv6 WG ml).
Please read below.
Cheers, Jan
-------- Original Message -------- Date: Tue, 19 Aug 2014 10:04:56 +0900 From: Seiichi Kawamura <kawamucho@mesh.ad.jp>
Fellow BCOPers
Hi there. Some folks in Japan, especially tech bloggers and tech documentation producers are saying that we need more ipv6 documentation prefix than just 2001:db8::/32
When describing a classic 3 prefix network topology they would use
net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48
where as with v4,
net A = 192.0.2.0/24 net B = 198.51.100.0/24 net C = 203.0.113.0/24
The 3 IPv6 prefixes are too similar and it's intuitively hard to tell if the 3 prefixes are talking about a network, or is it 3 separate networks. I guess this is bad especially for educational tutorial documentation.
So I'm thinking that if there are 2 more prefixes defined as documentation, I would say that's enough. We can maybe even revive 3ffe:: and make that documentation purpose.
However, I'm intersted in hearing opinions from other regions. Do you think there are any such needs in your region?
-Seiichi
_______________________________________________ Bcop-gc mailing list Bcop-gc@elists.isoc.org https://elists.isoc.org/mailman/listinfo/bcop-gc
OK. I joined the list so you don't have to cc me anymore. I noticed on a MyNOG presentation last week that someone was using 2001:db8:: to describe one network and 3fff:: to describe another. Not sure if he meant to use 3ffe:: but it certainly took me by surprise.
I thought this is what Japanese colleagues are suggesting...
To be exact, I am hearing these voices from the non-operator non-scientist people. The people who give basic tutorials, the tech bloggers, etc... -Seiichi (2014/08/22 23:27), Jan Zorz @ go6.si wrote:
On 22/08/14 02:56, 马严 wrote:
Hi, Jan and all,
As RFC3849 specified, the prefix reserved for documentation is a /32 block, 2001:DB8::/32 while people can use the following: net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48 we can also use net A = 2001:db8:1::/48 net B = 2001:db8:8000::/48 net C = 2001:db8:a000::/48 for being easy recognized as separated networks.
Yes, I agree, but this is different just to some limited extent. People, not very familiar with IPv6 and on their learning curve might mistakenly understand this as prefixes in one network. To be really sure they distinguish between the networks (being just different local networks or different AS-es) I think completely different IPv6 prefixes should be used, visually different from the first "character" on...
I thought this is what Japanese colleagues are suggesting...
(Including Seiichi-san to cc:)
Cheers, Jan
The only shortcoming that I can think of is, because 2001:db8::/32 is one big block, when being used to describe inter-domain network topology, /32 address block may easily be considered as all networks belong to one organization. Any comment?
I also cc:ed this email to the co-author of RFC3849, G.Huston, Chief Scientist from APNIC, for further discussion.
Best regards, --MA Yan
----- reply email ----- *Sender:*Jan Zorz @ go6.si <jan@go6.si> *Recipient:*bcop <bcop@ripe.net> *Time:*08/21/2014 22:11:55 *Subject:*[bcop] Fwd: [Bcop-gc] documentation ipv6 prefix
Dear RIPE BCOP community,
I got a question from Seiichi Kawamura, JANOG BCOP co-chair and I think this suggestion/question would be best if discussed here on this mailing list (and maybe also on IPv6 WG ml).
Please read below.
Cheers, Jan
-------- Original Message -------- Date: Tue, 19 Aug 2014 10:04:56 +0900 From: Seiichi Kawamura <kawamucho@mesh.ad.jp>
Fellow BCOPers
Hi there. Some folks in Japan, especially tech bloggers and tech documentation producers are saying that we need more ipv6 documentation prefix than just 2001:db8::/32
When describing a classic 3 prefix network topology they would use
net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48
where as with v4,
net A = 192.0.2.0/24 net B = 198.51.100.0/24 net C = 203.0.113.0/24
The 3 IPv6 prefixes are too similar and it's intuitively hard to tell if the 3 prefixes are talking about a network, or is it 3 separate networks. I guess this is bad especially for educational tutorial documentation.
So I'm thinking that if there are 2 more prefixes defined as documentation, I would say that's enough. We can maybe even revive 3ffe:: and make that documentation purpose.
However, I'm intersted in hearing opinions from other regions. Do you think there are any such needs in your region?
-Seiichi
_______________________________________________ Bcop-gc mailing list Bcop-gc@elists.isoc.org https://elists.isoc.org/mailman/listinfo/bcop-gc
On 25.08.14 16:21, "Seiichi Kawamura" <kawamucho@mesh.ad.jp> wrote:
OK. I joined the list so you don't have to cc me anymore. I noticed on a MyNOG presentation last week that someone was using 2001:db8:: to describe one network and 3fff:: to describe another. Not sure if he meant to use 3ffe:: but it certainly took me by surprise.
I personally think we must be careful to make documentation examples too different from the real world, as people tends to use examples as templates in their real network designs. So keeping just the 2001:db8::/32 for documenting examples for ones own network is more than enough. However, I do see the problem when an example shows connectivity between two different resource holders. IMHO, I think we should consider finding a method of differentiating different network in examples, either by assigning one or more /32 prefixes documentation prefixes, or allow the 2001:db8::/32 to become 2001:db8::/29. Changing RFC3849 to reflect the changes done in the RIPE region, where an LIR can get a /29 without any further documentation. This will allow for a more real world documentation both for large networks and multiple networks. I.E. NET A = 2001:db8::/32 Subnet a = 2001:db8:1::/48 Subnet b = 2001:db8:2::/48 Subnet c = 2001:db8:3::/48 NET B = 2001:dba::/32 Subnet a = 2001:dba:1::/48 Subnet b = 2001:dba:8000::/48 Subnet c = 2001:dba:a000::/48 NET C = 2001:dbf::/32 Subnet a = 2001:dba:1::/48 Subnet b = 2001:dba:10::/48 Subnet c = 2001:dba:20::/48 OR NET A = 2001:db8::/29 Subnet a = 2001:db8::/48 Subnet b = 2001:dba::/48 Subnet c = 2001:dbf::/48 BR Ragnar
On 27.08.14 10:54, "Anfinsen, Ragnar" <Ragnar.Anfinsen@altibox.no> wrote:
NET C = 2001:dbf::/32 Subnet a = 2001:dba:1::/48 Subnet b = 2001:dba:10::/48 Subnet c = 2001:dba:20::/48
Oops, typo, should be: NET C = 2001:dbf::/32 Subnet a = 2001:dbf:1::/48 Subnet b = 2001:dbf:10::/48 Subnet c = 2001:dbf:20::/48 BR Ragnar
Hi All these are really good points.
I personally think we must be careful to make documentation examples too different from the real world, as people tends to use examples as templates in their real network designs. So keeping just the 2001:db8::/32 for documenting examples for ones own network is more than enough.
I totally agree with this.
I personally think we must be careful to make documentation examples too different from the real world, as people tends to use examples as templates in their real network designs. So keeping just the 2001:db8::/32 for documenting examples for ones own network is more than enough.
Also strongly agree with this. And I think this is one reason why we see more 3ffe:: and fd00:: usage in documents. Thanks! -Seiichi (2014/08/27 17:54), Anfinsen, Ragnar wrote:
On 25.08.14 16:21, "Seiichi Kawamura" <kawamucho@mesh.ad.jp> wrote:
OK. I joined the list so you don't have to cc me anymore. I noticed on a MyNOG presentation last week that someone was using 2001:db8:: to describe one network and 3fff:: to describe another. Not sure if he meant to use 3ffe:: but it certainly took me by surprise.
I personally think we must be careful to make documentation examples too different from the real world, as people tends to use examples as templates in their real network designs. So keeping just the 2001:db8::/32 for documenting examples for ones own network is more than enough.
However, I do see the problem when an example shows connectivity between two different resource holders.
IMHO, I think we should consider finding a method of differentiating different network in examples, either by assigning one or more /32 prefixes documentation prefixes, or allow the 2001:db8::/32 to become 2001:db8::/29.
Changing RFC3849 to reflect the changes done in the RIPE region, where an LIR can get a /29 without any further documentation. This will allow for a more real world documentation both for large networks and multiple networks.
I.E.
NET A = 2001:db8::/32 Subnet a = 2001:db8:1::/48 Subnet b = 2001:db8:2::/48 Subnet c = 2001:db8:3::/48
NET B = 2001:dba::/32 Subnet a = 2001:dba:1::/48 Subnet b = 2001:dba:8000::/48 Subnet c = 2001:dba:a000::/48
NET C = 2001:dbf::/32 Subnet a = 2001:dba:1::/48 Subnet b = 2001:dba:10::/48 Subnet c = 2001:dba:20::/48
OR
NET A = 2001:db8::/29 Subnet a = 2001:db8::/48 Subnet b = 2001:dba::/48 Subnet c = 2001:dbf::/48
BR Ragnar
Hi Since I've been cc'ed here, and since you've asked, my personal opinion is that a /32 in IPv6 is perfectly capable of describing network scenarios that encompass some 4 billion network prefixes, assuming a 64 bit interface identifier space. If you are considering writing documentation that requires in excess of 4 billion distinct network prefixes than I would seriously suggest that you have problems far far greater than trying to map your intended example into a /32 prefix. :-) I personally see no rational basis for further documentation prefixes, and certainly no rational basis for reviving 3ffe. Geoff On 22 Aug 2014, at 10:56 am, 马严 <mayan@bupt.edu.cn> wrote:
Hi, Jan and all,
As RFC3849 specified, the prefix reserved for documentation is a /32 block, 2001:DB8::/32 while people can use the following: net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48 we can also use net A = 2001:db8:1::/48 net B = 2001:db8:8000::/48 net C = 2001:db8:a000::/48 for being easy recognized as separated networks. The only shortcoming that I can think of is, because 2001:db8::/32 is one big block, when being used to describe inter-domain network topology, /32 address block may easily be considered as all networks belong to one organization. Any comment?
I also cc:ed this email to the co-author of RFC3849, G.Huston, Chief Scientist from APNIC, for further discussion.
Best regards, --MA Yan
----- reply email ----- Sender:Jan Zorz @ go6.si <jan@go6.si> Recipient:bcop <bcop@ripe.net> Time:08/21/2014 22:11:55 Subject:[bcop] Fwd: [Bcop-gc] documentation ipv6 prefix
Dear RIPE BCOP community,
I got a question from Seiichi Kawamura, JANOG BCOP co-chair and I think this suggestion/question would be best if discussed here on this mailing list (and maybe also on IPv6 WG ml).
Please read below.
Cheers, Jan
-------- Original Message -------- Date: Tue, 19 Aug 2014 10:04:56 +0900 From: Seiichi Kawamura <kawamucho@mesh.ad.jp>
Fellow BCOPers
Hi there. Some folks in Japan, especially tech bloggers and tech documentation producers are saying that we need more ipv6 documentation prefix than just 2001:db8::/32
When describing a classic 3 prefix network topology they would use
net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48
where as with v4,
net A = 192.0.2.0/24 net B = 198.51.100.0/24 net C = 203.0.113.0/24
The 3 IPv6 prefixes are too similar and it's intuitively hard to tell if the 3 prefixes are talking about a network, or is it 3 separate networks. I guess this is bad especially for educational tutorial documentation.
So I'm thinking that if there are 2 more prefixes defined as documentation, I would say that's enough. We can maybe even revive 3ffe:: and make that documentation purpose.
However, I'm intersted in hearing opinions from other regions. Do you think there are any such needs in your region?
-Seiichi
_______________________________________________ Bcop-gc mailing list Bcop-gc@elists.isoc.org https://elists.isoc.org/mailman/listinfo/bcop-gc
Hi, I can understand both opinions, but prefer to keep one /32 as there should be enough bits for a clear organization assignment in documentations like - 2001:db8:axx::/40 - 2001:db8:bxx::/40 - 2001:db8:cxx::/40 ... that way there are one /40 for organization a, b and c if someone wouldn't like to call them 1, 2 or 3. It is also possible split above prefixes like - 2001:db8:a1xx::/40 - 2001:db8:a2xx::/40 - 2001:db8:b4xx::/40 - 2001:db8:b5xx::/40 ... For organization A region one and two + organization B region four and five where every region of an organization can still delegate 2^8 /48 to customers. Karsten Am Samstag, 23. August 2014, 06:11:07 schrieb Geoff Huston:
Hi
Since I've been cc'ed here, and since you've asked, my personal opinion is that a /32 in IPv6 is perfectly capable of describing network scenarios that encompass some 4 billion network prefixes, assuming a 64 bit interface identifier space. If you are considering writing documentation that requires in excess of 4 billion distinct network prefixes than I would seriously suggest that you have problems far far greater than trying to map your intended example into a /32 prefix. :-)
I personally see no rational basis for further documentation prefixes, and certainly no rational basis for reviving 3ffe.
Geoff
On 22 Aug 2014, at 10:56 am, ?? <mayan@bupt.edu.cn> wrote:
Hi, Jan and all,
As RFC3849 specified, the prefix reserved for documentation is a /32 block,
2001:DB8::/32
while people can use the following: net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48
we can also use
net A = 2001:db8:1::/48 net B = 2001:db8:8000::/48 net C = 2001:db8:a000::/48
for being easy recognized as separated networks. The only shortcoming that I can think of is, because 2001:db8::/32 is one big block, when being used to describe inter-domain network topology, /32 address block may easily be considered as all networks belong to one organization. Any comment?
I also cc:ed this email to the co-author of RFC3849, G.Huston, Chief Scientist from APNIC, for further discussion.
Best regards, --MA Yan
----- reply email ----- Sender:Jan Zorz @ go6.si <jan@go6.si> Recipient:bcop <bcop@ripe.net> Time:08/21/2014 22:11:55 Subject:[bcop] Fwd: [Bcop-gc] documentation ipv6 prefix
Dear RIPE BCOP community,
I got a question from Seiichi Kawamura, JANOG BCOP co-chair and I think this suggestion/question would be best if discussed here on this mailing list (and maybe also on IPv6 WG ml).
Please read below.
Cheers, Jan
-------- Original Message -------- Date: Tue, 19 Aug 2014 10:04:56 +0900 From: Seiichi Kawamura <kawamucho@mesh.ad.jp>
Fellow BCOPers
Hi there. Some folks in Japan, especially tech bloggers and tech documentation producers are saying that we need more ipv6 documentation prefix than just 2001:db8::/32
When describing a classic 3 prefix network topology they would use
net A = 2001:db8:1::/48 net B = 2001:db8:2::/48 net C = 2001:db8:3::/48
where as with v4,
net A = 192.0.2.0/24 net B = 198.51.100.0/24 net C = 203.0.113.0/24
The 3 IPv6 prefixes are too similar and it's intuitively hard to tell if the 3 prefixes are talking about a network, or is it 3 separate networks. I guess this is bad especially for educational tutorial documentation.
So I'm thinking that if there are 2 more prefixes defined as documentation, I would say that's enough. We can maybe even revive 3ffe:: and make that documentation purpose.
However, I'm intersted in hearing opinions from other regions. Do you think there are any such needs in your region?
-Seiichi
_______________________________________________ Bcop-gc mailing list Bcop-gc@elists.isoc.org https://elists.isoc.org/mailman/listinfo/bcop-gc
participants (6)
-
Anfinsen, Ragnar -
Geoff Huston -
Jan Zorz ＠ go6.si -
Karsten Thomann -
Seiichi Kawamura -
马严