August 2017 - db-wg - mailman.ripe.net

<auto-dbm@ripe.net> is slow this afternoon.
by RIPE Database Administration 17 May '18

17 May '18

Dear Colleagues, Because of a very high level of use this afternoon, the response time of <auto-dbm(a)ripe.net> is longer than normal. I have made contact with those who are responsible and I expect the situation to improve as the machine works through the queue. Please be patient. Obviously we at the RIPE NCC Database Group are continuously working to make the performance of the database as good as possible. However, we cannot make improvements in zero time. We thank you for your co-operation up to now and we ask for your continued co-operation so that everyone receives the best possible service. If you have anymore questions, please contact <ripe-dbm(a)ripe.net>. Regards, Ambrose Magee ____________________________ RIPE Database Administration.

2 2

Whois Master Database Patch
by Edward Shryane 31 Aug '17

31 Aug '17

Dear working group, Tomorrow afternoon, Friday 1 September, we will apply a patch to the Whois Master database server. This will cause a short downtime for Whois updates. Queries will not be affected. Regards, Ed Shryane RIPE NCC

1 0

Re: [db-wg] NWI-7: abuse-c implementation plan
by Tim Bruijnzeels 30 Aug '17

30 Aug '17

Dear working group, We have a proposed timeline for the implementation. Let me quote the proposal sent out on 28 June here, and comment in-line: > 1) Allow “abuse-c:” on INET(6)NUM and AUT-NUM objects > > We will deploy a new version of whois where the “abuse-c:” attribute will be allowed to exist on INET(6)NUM and AUT-NUM objects. The syntax will be the same as for the ORGANISATION object: > abuse-c: [optional] [single] [inverse key] > > The Abuse Finder logic will be updated to use the first "abuse-c:" found in this order of priority: > -the local "abuse-c:" in the resource object > -the "abuse-c:" in an ORGANISATION object referenced by the "org:" attribute in the resource object > -search up the resource object hierarchy (less specific objects) > --the "abuse-c:" in the less specific resource object > --the "abuse-c:" in an ORGANISATION object referenced by the "org:" attribute in the less specific resource object > > The search will stop when the top level resource object is reached. > > In addition to this we plan to introduce a business rule that will disallow “abuse-mailbox:” to be modified, or (re-)added to any object, other than ROLE objects. This will enable us to do a phased clean-up of this data without invalidating objects. We plan to have a whois release 1.90 for this (and the updated use of the ‘pn’ keyword mentioned under item #5 below) in the Release Candidate environment no later than Monday 2 October. We plan to deploy this version to production on Monday 16 October. > 2) Clean up of “abuse-mailbox:” on ORGANISATION objects > > Three cases exist: > > A) “abuse-mailbox:” is present and it is the same as the “abuse-mailbox:” in the also present “abuse-c:” reference (8919 cases) > > In this case we propose to simply delete the “abuse-mailbox:” attribute and inform the holder. We will work on a clear text for this email, but the main message would be: > "We removed “abuse-mailbox:” from your ORGANISATION object because this attribute is deprecated as the final step in the deployment of the "abuse-c:" attribute. However, we see that you were already using the preferred way to document abuse contact details for your organisation and you are referencing an “abuse-c:” ROLE object with the same email address in its its “abuse-mailbox:” attribute. No action is required.” > > > B) “abuse-mailbox:” is present and it is different from the “abuse-mailbox:” in the also present “abuse-c:” reference (1848 cases) > > Note that some of these cases have been caused by the fact that since the introduction of “abuse-c:” LIRs can no longer modify the value of the “abuse-mailbox:” attribute. The intent was that “abuse-mailbox:” would have been cleaned up earlier, but unfortunately that dit not happen. That being said “abuse-c:” has been kept up to date by the LIR, and out of the two is also the one that is being shown on the Abuse Finder. > > Therefore we propose to delete the “abuse-mailbox:” attribute in these cases, and send a different notification to the holder with the following main message: > > "We removed “abuse-mailbox:” with value $email from your ORGANISATION object because this attribute is deprecated as the final step in the deployment of the "abuse-c:" attribute. However, we see that you were already using the preferred way to document abuse contact details for your organisation and you are referencing an “abuse-c:” ROLE object with that uses $email_2 in its “abuse-mailbox:” attribute. If the latter is correct no action is required. However if you want to modify the email address in your “abuse-c:” ROLE object, you can do so here (link to webupdates).” > > > C) “abuse-mailbox:” is present, and there is no “abuse-c:” reference (28312 cases) > > We *could* create “abuse-c:” ROLE objects for these cases, using the same maintainers as the ORGANISATION object. However, the “abuse-mailbox:” attribute has not been shown in the Abuse Finder, and this change could lead to wrong or out-of-date addresses suddenly receiving abuse reports. Therefore we suggest that we delete the attribute instead, and send the following main message to the holder of the ORGANISATION object: > > "We removed “abuse-mailbox:” with value $email from your ORGANISATION object because this attribute is deprecated as the final step in the deployment of the "abuse-c:" attribute. However, if you wish to add an email address for abuse reports you can do so by editing your object here (link to web updates for the ORGANISATION object)." > > We plan to perform these clean-ups on Monday 30 October. > 3) Clean up of “abuse-mailbox:” on PERSON (89k cases), MNTNER (2500 cases) and IRT (238 cases) > > In these cases we propose to send a warning email to the holders of these objects with the following main message: > > “The “abuse-mailbox:” is deprecated on these object and will be removed in 4 weeks, if you want to set up abuse contact information on your resource objects you can do so by having a default “abuse-c:” on your ORGANISATION referenced by your resource objects, or possibly overriding that default with a specific “abuse-c:” reference on applicable resource objects." > > And then 4 weeks later we will preform the clean-up and refer back to this communication. On further reflection we would prefer to alter the plan, and just a single clean-up of these objects on Monday 30 October as well. This way the clean-up process will be consistent across all object types. Furthermore, there is no action to be taken on these objects and “abuse-mailbox:” on these objects is ignored in the Abuse Finder. In short there does not seem to be a strong motivation for bothering the holders of these objects twice. > 4) Clean up whois code > > Now that “abuse-mailbox:” only appears where it should, we will be able to update the whois code itself so that “abuse-mailbox:” is no longer syntactically allowed on objects other than ROLE objects. This will allow us to remove the business rule added in phase 1 and simplify the whois code and maintenance. We will also update the RIPE Database Documentation to reflect this. We plan to do this as part of a future whois 1.91 release, as yet unplanned. This clean-up will not change any publicly visible functionality. > > > 5) Inverse queries > > As requested we can change the behaviour of the “pn” short cut to also search for “abuse-c:” in inverse queries. I.e. one can use: > whois -r -i pn uk41-ripe > > Instead of: > whois -r -i abuse-c,admin-c,tech-c,zone-c,author uk41-ripe As mentioned above, we plan to include this in the whois 1.90 release. Kind regards Tim Bruijnzeels Assistant Manager Software Engineering RIPE NCC Database Group > On 4 Aug 2017, at 14:39, William Sylvester via db-wg <db-wg(a)ripe.net> wrote: > > > From: William Sylvester <william.sylvester(a)addrex.net> > Subject: NWI-7: abuse-c implementation plan > Date: 3 August 2017 at 21:12:06 GMT+2 > To: "db-wg(a)ripe.net" <db-wg(a)ripe.net> > > > Working group members, > > Based on the responses we have received for NWI-7 : Abuse-c implementation, there is a clear consensus for the proposed implementation that has been presented by Tim from RIPE NCC as written for improving "abuse-c:". > > At this time, the DB-WG co-chairs would like to ask the RIPE NCC to proceed with this implementation. > > Kind Regards, > > DB-WG Chairs > > > > >

1 0

Re: [db-wg] out of region routing in the RIPE Database
by Clement Cavadore 14 Aug '17

14 Aug '17

On Tue, 2017-07-18 at 15:26 +0200, Nick Hilliard via db-wg wrote: > > I am not in favour of having the RIPE database as an open-access > database on the basis that this mixes up two sets of data, > authoritative > and non-authoritative, and it it is impossible for someone casually > querying the database to determine which is which. > > Some people are inserting random route: objects into the database, and > those route: objects are being picked up by provisioning systems and > ending up configured on routers and IXP route servers. This enables > prefix hijacking, which is a pressing operational issue. I agree with Nick's position. It legitimates what seems to be rogue announcements, like for example 196.16.0.0/14, as mentionned recently on the NANOG mailing list (*). We should, IMHO not be able to insert out of region route(6) object without having a prior authentication mechanism, or making it be specially flagged, so the auto ACL system from upstreams wouldnt match it. (*) https://mailman.nanog.org/pipermail/nanog/2017-August/091954.html -- Clément Cavadore

1 0

Re: [db-wg] out of region routing in the RIPE Database
by Daniel Shaw 11 Aug '17

11 Aug '17

On 11/08/2017, 05:14, Randy Bush via db-wg typed: > >> >> databases. It's that they never have, and it's just the "way things >> have been done". It's always ended up that they *can* support multiple >> databases if asked to. > > when negotiating with a prospective peer, i would rather not add "you > will have to change operational practice and code for this to work." That is a good point. I would also rather not, generally. And that is the basis of my original fears around this something like a year ago. What I found in practise (so far) has been that when stated not as "you will *have* to change operations", but as "would it be possible for you to change operations for me?", the answer has usually been "Ok". I have to be honest, that was a surprise. - Daniel

1 0

Re: [db-wg] out of region routing in the RIPE Database
by Randy Bush 11 Aug '17

11 Aug '17

> That has proven to never really be an issue. When a network upstream > of another network says they don't support additional databases for > IRR, so far that has always proven to actually mean they've never > needed to think about supporting multiple databases, and never had the > demand to do the work to change that. > > In other words, it's not that they *can't* support other > databases. It's that they never have, and it's just the "way things > have been done". It's always ended up that they *can* support multiple > databases if asked to. when negotiating with a prospective peer, i would rather not add "you will have to change operational practice and code for this to work." randy

1 0

Re: [db-wg] out of region routing in the RIPE Database
by Daniel Shaw 10 Aug '17

10 Aug '17

Hi there, On 09/08/2017, 19:34, Stefan Ideler via db-wg typed: > > > As a result our route objects regardless of origin are noted in the RIPE database. Our upstreams use it to generate their prefix filters (and not all of them support multiple databases) I won't comment here on your specific use case or the overall reasoning or motivation for wanting to use the RIPE DB exclusively, even for resources that may come from various other regions. I have an opinion, but it's just that. An opinion. What I'd like to comment on is the excerpt above. Specifically about upstream providers not supporting multiple databases. That has proven to never really be an issue. When a network upstream of another network says they don't support additional databases for IRR, so far that has always proven to actually mean they've never needed to think about supporting multiple databases, and never had the demand to do the work to change that. In other words, it's not that they *can't* support other databases. It's that they never have, and it's just the "way things have been done". It's always ended up that they *can* support multiple databases if asked to. I speak from experience here. If you go back into the archives of this list, when this issue was first brought to the list, I responded with this exact fear: But what about the upstream support? Since then I've had the opportunity to speak to folks from some of the larger networks that are often someone's upstream on behalf of AFRINIC members from time to time, and also in my own capacity in the context of an RIR meeting event network. We've also encouraged member operators to just ask their upstreams to change. And this has proved to never have remained an issue to the best of my knowledge. Summary: While you may (or may not) have other good reasons for the RIPE DB to be open (or not), upstream support of alternates is not likely one of the reasons. Best regards, Daniel

1 0

Re: [db-wg] out of region routing in the RIPE Database
by Nick Hilliard 09 Aug '17

09 Aug '17

Stefan Ideler via db-wg wrote: > Being a member of multiple RIR's, the Ripe db is user friendly and a > scriptable database. I don't want to divide my as-sets and prefix > filter generation over 3 or more different database systems, with > the additional maintenance and in some cases even less verification. > Atleast my RIPE originating objects are secure and they make up the > majority of my prefixes. > > As a result our route objects regardless of origin are noted in the > RIPE database. Our upstreams use it to generate their prefix filters > (and not all of them support multiple databases) and as a result we > require all our downstream customers to have their AS/route objects > within the RIPE database as well. Several years ago, we planned this sort of approach via IXP Manager, but when we wrote the code to do it, we realised that this was a very limited and limiting mechanism for handling prefix list generation. It caused a lot of extra work for us in terms of support requirements and explaining to people that they had to do this. Several member organisations refused to comply and said they had no intention of changing their policies. The result of this was that we ended up not carrying traffic that we could alternatively have carried, and we annoyed our members. Not the best idea, in retrospect. We ended up having to be a lot more flexible on the issue. The best way we found to handle it was to allow any customer to be associated with any arbitrary IRR policy. E.g. one customer could use source: RIPE, ARIN and RADB, and another could use LACNIC, ALTDB and LEVEL3. It's actually pretty straightforward to implement this, and gives the advantage that you're no longer creating extra work for customers by forcing them to register their route: objects in IRRDBs where they wouldn't otherwise do so. The code for this is freely available on github.com/inex/IXP-Manager. I'm mentioning this because there is a serious security problem associated with permitting out-of-region objects in the RIPE database: this causes repeated, serious security problems with routing leakage on the Internet. This feature is being persistently abused by scammers and other criminals and is devaluing the integrity of the data in the RIPE IRRDB. Unless this problem is fixed, it's going to continue to be a problem indefinitely. We cannot fix it unless people fix their code. Nick

1 0

Re: out of region routing in the RIPE Database
by Stefan Ideler 09 Aug '17

09 Aug '17

Hi, > I would like to have the IRR kept open, we have space from various > regions and need to be able to make use of a single well known IRR in > order to register our policies and be able to mix our ASNs with our > space. >This might indeed be the way forward for "I have a customer in my RIPE >AS which has APNIC space and wants to document that in a way that that >provisioning tools can pick this with sufficient trust on data quality". I agree with the above. Being a member of multiple RIR's, the Ripe db is user friendly and a scriptable database. I don't want to divide my as-sets and prefix filter generation over 3 or more different database systems, with the additional maintenance and in some cases even less verification. Atleast my RIPE originating objects are secure and they make up the majority of my prefixes. As a result our route objects regardless of origin are noted in the RIPE database. Our upstreams use it to generate their prefix filters (and not all of them support multiple databases) and as a result we require all our downstream customers to have their AS/route objects within the RIPE database as well. As such I oppose removing the functionality to register "foreign route objects" and instead support that a proper verification solution between the different RIR's should be implemented (some examples described in this thread). Regards, Stefan Sent via RIPE Forum -- https://www.ripe.net/participate/mail/forum

1 0

Re: [db-wg] NWI-7: abuse-c implementation plan
by Tim Bruijnzeels 07 Aug '17

07 Aug '17

Dear working group, We will come back with a more detailed implementation time line later this week. Informally speaking we expect that we can do the technical work to prepare the phase-1 whois release in the coming weeks. We will probably have to plan the clean-up after the holidays. But, we will get back to you once we have a more definite plan. Kind regards, Tim Bruijnzeels Assistant Manager Software Engineering RIPE NCC Database Group > On 4 Aug 2017, at 14:39, William Sylvester via db-wg <db-wg(a)ripe.net> wrote: > > > From: William Sylvester <william.sylvester(a)addrex.net> > Subject: NWI-7: abuse-c implementation plan > Date: 3 August 2017 at 21:12:06 GMT+2 > To: "db-wg(a)ripe.net" <db-wg(a)ripe.net> > > > Working group members, > > Based on the responses we have received for NWI-7 : Abuse-c implementation, there is a clear consensus for the proposed implementation that has been presented by Tim from RIPE NCC as written for improving "abuse-c:". > > At this time, the DB-WG co-chairs would like to ask the RIPE NCC to proceed with this implementation. > > Kind Regards, > > DB-WG Chairs > > > > >

1 0