- db-wg - mailman.ripe.net

Summary of the Migration issues
by Andrei Robachevsky 20 Mar '01

20 Mar '01

Dear Colleagues, Please let me summarize the migration issues that have been discussed already and highlight some issues that require more attention. Best regards, Andrei Robachevsky DB Group Manager RIPE NCC ---------------------------------------- 1. Route object creation (RFC2725 implications) In order to create route object in the RIPE DB, one needs to have corresponding objects (inetnum or less specific route, and aut-num) and pass proper authorization from these objects. The problem occurs when someone needs to register route object which has either prefix, or origin, or the both from non RIPE IP/ASN space. Proposed solution: 1.1 An as-block object(s) encompassing the non RIPE ASN space will be created with "mnt-lower:" protected by mntner with NONE auth. 1.2 An encompassing inetnum object with "mnt-routes:" protected by mntner with NONE auth will be created. This will eliminate need for corresponding inetnum creation and will reduce amount of redundant information. 1.3 In case origin of a route object being created is from non RIPE space, users will need to create corresponding aut-num object in the RIPE DB. 2. Updates in RIPE-181 format Migration plan allows users to send updates in RIPE-181 format for some time (until May 14th to <auto-dbm(a)ripe.net>, from May 14th till October 15th to <auto-181(a)ripe.net>) after the migration. In such case objects are automatically converted to RPSL. However the switchover to the new version of the RIPE Database occurs on 23 of April, and since then the following constraints will apply: 2.1 PGP authentication PGP authorization scheme will not work with <auto-dbm(a)ripe.net> robot after 23 April. Updates requiring PGP auth should be sent directly to <auto-rpsl(a)ripe.net>. The problem occurs because after ripe181->RPSL transformation it is impossible to verify signature. For this reason, some updates requiring PGP auth may be rejected during migration. It advisable not to send such updates from 22 April, 9a.m. till 23 April 9a.m. 2.2 New syntax rules When processing an object after ripe181->RPSL conversion new v3.0 syntax rules apply. The results of this will be: - empty attributes (i.e. having no value) are not allowed; - some attributes that were optional before are now mandatory; Also some undocumented behaviour of the v2.x Database will not be supported, namely: - attribute aliases (please see list of aliases attached) are not allowed; - inetnum object cannot be specified in prefix notation. 2.3 New authorization rules New authorization rules defined by RFC2725 apply to creation of aut-num and route objects. See 1. 3. Objects with names containing reserved RPSL prefixes They will be renamed on 2 of April. All references will be updated accordingly. Currently there are 6 of them in the RIPE Database. mntner: AS-5532MNTB mntner: AS-NETLINK-MNT mntner: AS-THENET-MNT mntner: AS-NEW1-MNT mntner: RS-MNT mntner: AS-MNT 4. Broken PGP certificates. They will be deleted on 2 of April. ---------------------- v2.x attribute aliases (not supported in v3.0!): Alias Attrbute change -> changed email -> e-mail fax -> fax-no remark -> remarks deleted -> delete asname -> as-name rev-svr -> rev-srv network -> inetnum netnum -> inetnum

3 3

RFC2725
by Holger Muenx 15 Mar '01

15 Mar '01

Guten Tag, I hope this is the correct forum - if not please direct me to a more appropriate one. After following the discussion for a while I am still unsure about the exact consequences of the activation of RFC2725 on 23-Apr-2001. I will try to summarize my understanding: Please check this and correct me where necessary. You will find a list of open questions at the bottom of this mail. As Frank Bohnsack wrote in http://www.ripe.net/ripe/mail-archives/db-wg/current/msg00090.html AS numbers are allocated to regional registries as shown in http://www.isi.edu/in-notes/iana/assignments/as-numbers and the IPv4 address space allocated to various registries as shown in http://www.isi.edu/in-notes/iana/assignments/ipv4-address-space RFC2725 states in section 9.9 that during addition of an inetnum the most specific matching inetnum is being searched to find the required authentication. During addition of a route the aut-num of the originating AS and either a most specific matching inetnum or a most specific matching route is being searched to find the required authentication. This leaves UUNET in a curious situation. For example, UUNET uses AS 702 which is in the range 1 - 1876 allocated by the ARIN and the inetnum 193.96.0.0/13 which is in 193.0.0.0/8 allocated by the RIPE. With RFC2725 in place UUNET would have its aut-num record in a different registry than its inetnum, rendering it impossible to add routes for the 193.0.0.0/8. There are probably more complications caused by moving data from one registry to another. For example, some people will continue to expect information at the original registry because they did not prepare for RFC2725, breaking mechanisms such as automatic prefix filter creation. Andrei Robachevsky presented two suggestions in http://www.ripe.net/ripe/mail-archives/db-wg/current/msg00095.html Suggestion 1 means that RIPE will implement changes so that RFC2725 will be in effect for AS numbers and IP address space allocated by RIPE only. No checks according to RFC2725 will be done for all other AS numbers and IP address space. Suggestion 2 means that there will be encompassing inetnum objects which contain the address ranges assigned to other registries than RIPE. These objects will have a maintainer with NONE authentication so that everybody can add routes within these inetnum which refer to their own aut-num records. No decision has been made on whether one of the suggestion and, if so, which of them will be implemented. So far my understanding. Please review - if it is completely wrong I hope that I at least amused you. ;-) Thank you for your help! Holger Münx EMEA Access & Backbone Networks UUNET Deutschland GmbH Tel. +49 231 972 0 Sebrathweg 20 Fax. +49 231 972 1111 44149 Dortmund, Germany Holger.Muenx(a)de.uu.net http://www.de.uu.net/

1 0

RE: RIPE DB transition and RFC2725
by Lu, Ping 09 Mar '01

09 Mar '01

Comments followed. Ping Lu Cable & Wireless Global Network Tools and Analysis Group, USA W: +1-703-292-2359 E: plu(a)cw.net > -----Original Message----- > From: Cengiz Alaettinoglu [mailto:cengiz@packetdesign.com] > Sent: Friday, March 09, 2001 11:24 AM > To: Lu, Ping > Cc: 'Frank.Bohnsack(a)de.uu.net'; db-wg(a)ripe.net > Subject: RE: RIPE DB transition and RFC2725 > > > > Lu, Ping (plu(a)cw.net) on March 9: > > Finally somebody express the same concerns as ours. > > > > I have a discussion with the RIPE staff at the RIPE-38 > meeting about this. > > > > There are two seperate issues here: > > 1. The RFC doesn't say a implementation (like RIPE-DB v3.0) > should allow > > foreign references ( using objects from other RR database) or not. > > RFC does not prohibit foreign references. As a matter of fact, RIPE > has a workaround that does this. > > Please also see the companion RFC RFC 2769 for making RFC 2725 checks > across registries. RFC 2769 and 2725 together yield a global > distributed registry. Of course, it requires cooperation form other > (at lease regional) registries. > The wrokaround means DISABLE integrity checking. That's an option for RIPE-DB v2.x (RIPE-181) but we haven't tried RIPE-DB v3.x yet. Also the workaround raises another question: How do you authorize the route creation from a foreign AS ? That's exactly the issue from Andrei's previous email (subject: Migration issues: route creation). And the solution to this needs to be discussed further. ( Hi, Andrei, any more suggestions from the community yet ?) > > 2. The assumption of RIPE-DB v3.0 is that all object > references have to be > > resloved locally( ie. with source: RIPE). > > > > The 2nd issue means if any ISP with ARIN's AS number want > to register a > > route with RIPE you have to register your PN(person) then > your MT(mntner) > > then your AN(aut-num) with RIPE first. > > > > So each ISP has to maintain a different set of objects with each RR. > > AS123(in RIPE) -> MNT-AS123-RIPE -> NIC456-RIPE > > AS123(in ARIN) -> MNT-AS123-ARIN -> NIC456-ARIN > > AS123(in APNIC) -> MNT-AS123-APNIC -> NIC456-APNIC > > AS123(in JP) -> MNT-AS123-JP -> NIC456-JP > > AS123(in DE) -> MNT-AS123-DE -> NIC456-DE > > .... > > > > Just to remember to update all these information everytime > there is a little > > tiny change. > > It is fun, isn't it ? > > > > And NO you can't get all route information from just one RR > unless we all > > sit down and seriously > > talking about the GLOBAL issues. > > > > Ping Lu > > Cable & Wireless Global > > Network Tools and Analysis Group, USA > > W: +1-703-292-2359 > > E: plu(a)cw.net > > > > > > > -----Original Message----- > > > From: Frank Bohnsack [mailto:Frank.Bohnsack@de.uu.net] > > > Sent: Friday, March 09, 2001 7:43 AM > > > To: db-wg(a)ripe.net > > > Subject: RIPE DB transition and RFC2725 > > > > > > > > > Dear colleagues, > > > > > > Yes, I know the deadline for transition comments is over, but > > > after studying RFC2725 we need a confirmation for our > understanding. > > > > > > RFC2725 defined the need for an "as-block" object for each related > > > "aut-num" object and an "inetnum" object for each related > > > "route" object. > > > > > > The assignment of AS blocks and IP space is managed by IANA. > > > * http://www.isi.edu/in-notes/iana/assignments/as-numbers > > > * http://www.isi.edu/in-notes/iana/assignments/ipv4-address-space > > > > > > Does the new RIPE DB manage only "as-blocks" and > "inetnums" which are > > > assigned to RIPE ? > > > > > > 1/ And if I'm right, is it true that we can't store our ASes > > > 702, 1270 > > > and a lot of routes (assigned to ARIN) in the new > RIPE database ? > > > > > > 2/ And is it true that we therefore can't set our route > > > (assigned to RIPE) > > > objects origin to our ASes ? > > > > > > The solution for 1/ is moving to ARIN, but what might be the > > > solution for 2/ > > > ? > > > > > > You say a big ISP is anyway talking to multiple IRRs, but I > > > think customers > > > who want to peer with these can't get complete information > > > while using RIPE > > > DB only. > > > > > > > > > Cheers > > > Frank > > > > > > -- > > > Frank Bohnsack email fb(a)de.uu.net > > > UUNET, A Worldcom Company phone +49 (0)231 972-1495 > > > EMEA Access & Backbone Networks fax +49 (0)231 972-1188 > > > Team Dortmund web www.de.uu.net > > > > > > Cengiz > > -- > Cengiz Alaettinoglu Packet Design >

2 1

RE: RIPE DB transition and RFC2725
by Lu, Ping 09 Mar '01

09 Mar '01

Finally somebody express the same concerns as ours. I have a discussion with the RIPE staff at the RIPE-38 meeting about this. There are two seperate issues here: 1. The RFC doesn't say a implementation (like RIPE-DB v3.0) should allow foreign references ( using objects from other RR database) or not. 2. The assumption of RIPE-DB v3.0 is that all object references have to be resloved locally( ie. with source: RIPE). The 2nd issue means if any ISP with ARIN's AS number want to register a route with RIPE you have to register your PN(person) then your MT(mntner) then your AN(aut-num) with RIPE first. So each ISP has to maintain a different set of objects with each RR. AS123(in RIPE) -> MNT-AS123-RIPE -> NIC456-RIPE AS123(in ARIN) -> MNT-AS123-ARIN -> NIC456-ARIN AS123(in APNIC) -> MNT-AS123-APNIC -> NIC456-APNIC AS123(in JP) -> MNT-AS123-JP -> NIC456-JP AS123(in DE) -> MNT-AS123-DE -> NIC456-DE .... Just to remember to update all these information everytime there is a little tiny change. It is fun, isn't it ? And NO you can't get all route information from just one RR unless we all sit down and seriously talking about the GLOBAL issues. Ping Lu Cable & Wireless Global Network Tools and Analysis Group, USA W: +1-703-292-2359 E: plu(a)cw.net > -----Original Message----- > From: Frank Bohnsack [mailto:Frank.Bohnsack@de.uu.net] > Sent: Friday, March 09, 2001 7:43 AM > To: db-wg(a)ripe.net > Subject: RIPE DB transition and RFC2725 > > > Dear colleagues, > > Yes, I know the deadline for transition comments is over, but > after studying RFC2725 we need a confirmation for our understanding. > > RFC2725 defined the need for an "as-block" object for each related > "aut-num" object and an "inetnum" object for each related > "route" object. > > The assignment of AS blocks and IP space is managed by IANA. > * http://www.isi.edu/in-notes/iana/assignments/as-numbers > * http://www.isi.edu/in-notes/iana/assignments/ipv4-address-space > > Does the new RIPE DB manage only "as-blocks" and "inetnums" which are > assigned to RIPE ? > > 1/ And if I'm right, is it true that we can't store our ASes > 702, 1270 > and a lot of routes (assigned to ARIN) in the new RIPE database ? > > 2/ And is it true that we therefore can't set our route > (assigned to RIPE) > objects origin to our ASes ? > > The solution for 1/ is moving to ARIN, but what might be the > solution for 2/ > ? > > You say a big ISP is anyway talking to multiple IRRs, but I > think customers > who want to peer with these can't get complete information > while using RIPE > DB only. > > > Cheers > Frank > > -- > Frank Bohnsack email fb(a)de.uu.net > UUNET, A Worldcom Company phone +49 (0)231 972-1495 > EMEA Access & Backbone Networks fax +49 (0)231 972-1188 > Team Dortmund web www.de.uu.net >

3 2

RIPE DB transition and RFC2725
by Frank Bohnsack 09 Mar '01

09 Mar '01

Dear colleagues, Yes, I know the deadline for transition comments is over, but after studying RFC2725 we need a confirmation for our understanding. RFC2725 defined the need for an "as-block" object for each related "aut-num" object and an "inetnum" object for each related "route" object. The assignment of AS blocks and IP space is managed by IANA. * http://www.isi.edu/in-notes/iana/assignments/as-numbers * http://www.isi.edu/in-notes/iana/assignments/ipv4-address-space Does the new RIPE DB manage only "as-blocks" and "inetnums" which are assigned to RIPE ? 1/ And if I'm right, is it true that we can't store our ASes 702, 1270 and a lot of routes (assigned to ARIN) in the new RIPE database ? 2/ And is it true that we therefore can't set our route (assigned to RIPE) objects origin to our ASes ? The solution for 1/ is moving to ARIN, but what might be the solution for 2/ ? You say a big ISP is anyway talking to multiple IRRs, but I think customers who want to peer with these can't get complete information while using RIPE DB only. Cheers Frank -- Frank Bohnsack email fb(a)de.uu.net UUNET, A Worldcom Company phone +49 (0)231 972-1495 EMEA Access & Backbone Networks fax +49 (0)231 972-1188 Team Dortmund web www.de.uu.net

1 0

Re: Announcement: Hardware upgrade of whois server
by Stephen Burley 07 Mar '01

07 Mar '01

Sorry for any duplication > Dear colleagues, > > After experiencing performance slow-down on our whois service, we > decided > to upgrade the hardware of our primary whois server. > > The upgrade will take place on 9 March, 2001, Friday. It will start at > 8:00am CET and planned to last approximately 2 hours. Just a quick quaestion, Why Friday and why during working hours? All our maint slots have to be out of hours and not preceding a weekend. Curious -- ------------------------------------------------------------------------ Stephen Burley "If patience is a virtue, and ignorance is bliss, UUNET EMEA Hostmaster you can have a pretty good life [SB855-RIPE] if you're stupid and willing to wait" ------------------------------------------------------------------------

2 1

Announcement: Hardware upgrade of whois server
by RIPE Database Administration 07 Mar '01

07 Mar '01

Dear colleagues, After experiencing performance slow-down on our whois service, we decided to upgrade the hardware of our primary whois server. The upgrade will take place on 9 March, 2001, Friday. It will start at 8:00am CET and planned to last approximately 2 hours. During the operation, whois queries will be directed to our secondary server and the incoming update messages will be queued, to be processed later. We apologize for any inconvenience this may cause. Regards, Engin Gunduz ____________________________ RIPE Database Administration.

2 1

RIPE RPSL transition date set!
by SchmitzJo＠aol.com 05 Mar '01

05 Mar '01

Hello all, following up on our last call for comments on the RPSL transition for the RIPE database, we did not receive any mails or information which may have delayed the transition. In contrast, we received a set of supportive mails. Accordingly, the transition to RPSL will now follow the outlined dates: o 23-Apr-2001: switching to the RPSL database *** now set *** - All objects are automatically converted to RPSL format - Queries return RPSL only - Mirroring follows new format and rules at whois.ripe.net, port 4444 - RPSL formatted updates can be submitted to auto-rpsl(a)ripe.net - RIPE-181 updates are still accepted at auto-dbm(a)ripe.net, but are autmatically converted to RPSL - RFC2725 (Routing Policy Security System) rules apply throughout o 14-May-2001: exchanging mail addresses for submitting RIPE181 and RPSL-formatted updates - auto-dbm(a)ripe.net only accepts RPSL updates - RIPE-181 updates are still possible but now go to auto-181(a)ripe.net and are automatically converted to RPSL (again, RFC2725 rules apply here as well) o 15-Oct-2001: RIPE-181 updates are no longer accepted For further reading, please refer to the migration web page http://www.ripe.net/rpsl We will continue to update you prior to the dates detailed above. Thanks - the RIPE NCC and the chairmen of Routing: Joachim Schmitz DB: Wilfried Woeber LIR: Hans Peter Holen

1 2

Re: RIPE: last call for comments RPSL transition
by SchmitzJo＠aol.com 02 Mar '01

02 Mar '01

Dear Christian, you wrote: > well I'd say go for it. Thank you for your support. > But let me kindly ask for the following thingies: > > since the RIPE-NCC is now in charge of the maintenance of the RAtoolset > please make sure that there's a release available (specifically > RTconfig) which is fully compatible with the RIPE RPSL server by mid of > March 2001 latest. I believe that this request is more than legitimate, and that the NCC is hearing you very well - after all it is in all our interest to have this running. > > It appears that this might be available already (since Feb20 ?), > however, the link to > > "RtConfig Patch for the RAToolSet" of the "RPSL Migration" homepage is > broken. This is currently worked on, as I hear. > > In addition, any clear links to concise How2do documents on RAToolSet > and RtConfig would be appreciated. On the one hand, I think the best source for that still is http://www.isi.edu/ra/RAToolSet/ On the other hand, we will have a tutorial on the RAToolSet during RIPE39. This has been confirmed by now, and I believe that additional documentation will become available around it. Thanks Joachim --- JS395-RIPE -- standard disclaimer ---

1 0

Re: RIPE: last call for comments RPSL transition
by SchmitzJo＠aol.com 02 Mar '01

02 Mar '01

Dear Ping Lu, you wrote: > We had a lively discussion about the RPSL transition during the RIPE-38 > meeting. This is correct, and I very much appreciate the input you provided during our joint working group session. > Well, if we can't fight them then the best thing to do is to join them. Do I need to be worried now? I never considered this to be a fight ;-) > With the company's support behind us, we (Cable & Wireless Global) as a > legacy LIR (used to be MCI) > would like to work with RIPE to help the RPSL transition. > > CW will adopt a similar schedule as RIPE's to transform our Routing Registry > into the new RPSL format. > This will probably make all the mirroring LIRs' job much easier and > hopefully all the downstream ISPs happier (only the RPSL format to deal > with). > > So GO FOR IT. I welcome your approval. I believe I do not exaggerate stating that you can count on the support from the RIPE community and the NCC in particular in the transition effort. Thanks Joachim --- JS395-RIPE -- standard disclaimer ---

1 0