- db-wg - mailman.ripe.net

Re: From Address Forged
by Michael van Elst 11 Nov '98

11 Nov '98

On Thu, Nov 05, 1998 at 06:01:23PM +0100, Siegfried Langenbach wrote: > On 5 Nov 98, at 17:45, Daniel Karrenberg wrote: > > > > > Wilhelm and other colleagues, > > > > it appears that the header forging did only occur in an effort to > > correct mistakes made earlier. Repeat: the updates that caused the > > problems originally were not done using forged headers. This does not > > mean header forging was not very bad judgement, but at least it occurred > > only in an effort to correct earlier mistakes. The RIPE NCC will work > > with Transpac on a post-mortem analysis. More news as it develops. > > Hmm, > if it was needed to forge a header to correct a mistake, > how then the mistake could be done without forging the header? > > siegfried I get tired discussing this. Yes, the original 'mistake' didn't involve forged headers. Yes, old records have been deleted this way. No, these records were not protected by their maintainers (auth: NONE). Yes, the handles were reallocated to other records. Yes, some of these new records were protected by their maintainers. Yes, these new records were deleted using forged mail headers. No, the old records were not restored because the restoring update was sent too early (usually in the same mail which inherently fails). Yes, the yet-again free handles have been reallocated to even more new records again. Yes, deleting the new records caused more inconsistencies and more data loss. No, deleting the new records didn't correct the original 'mistake' at all. Yes, deleting the new records produced more errors outside the RIPE-DB realm (for example, registrations of domains that reference the new-but-then-deleted person records in the RIPE-DB failed). Yes, deleting the new records sabotaged other people's attempt to recover from original 'mistake'. We have thus switched from MAIL-FROM to CRYPT-PW authorization. If I see any unauthorized update in the future, the responsible person gets shot. -- i.A. Michael van Elst / phone: +49 721 6635 330 Xlink - Network Information Centre \/ fax: +49 721 6635 349 Vincenz-Priessnitz-Str. 3 /\ link http://nic.xlink.net/ D-76131 Karlsruhe, Germany /_______ email: hostmaster(a)xlink.net [ Xlink Internet Consulting GmbH, Sitz Koeln ] [ Amtsgericht Koeln HRB 3526, Geschaeftsfuehrer: Michael Rotert ]

2 2

RE: From Address Forged
by Winfried Haug 05 Nov '98

05 Nov '98

First of all, please dont quote complete mails for just 2 lines from you.. > Hmm, > if it was needed to forge a header to correct a mistake, > how then the mistake could be done without forging the header? > if you have attended at this topic you should have noticed that most handles had no protection (just a notify) but were not mounted. So there was no problem to delete these records... Thats it.. Winfried

1 0

Re: From Address Forged
by Daniel Karrenberg 05 Nov '98

05 Nov '98

Wilhelm and other colleagues, it appears that the header forging did only occur in an effort to correct mistakes made earlier. Repeat: the updates that caused the problems originally were not done using forged headers. This does not mean header forging was not very bad judgement, but at least it occurred only in an effort to correct earlier mistakes. The RIPE NCC will work with Transpac on a post-mortem analysis. More news as it develops. Daniel

1 0

Réf. : From Address Forged
by GALL Thierry TPC-SRD 04 Nov '98

04 Nov '98

Actually, the address was "forged" but only partially by Trevor, and Hostmaster was in copy of the mail, More over it was an attempt to repair the original error and to avoid the work necessary to rebuild all the links that have destroyed. Please apologize again for all these perturbations. Thierry GALL

1 0

Re: Events of Tuesday, 27 October 1998.
by Havard.Eidnes＠runit.sintef.no 03 Nov '98

03 Nov '98

Dimitrios Kaloger wrote: > Dear Madam,/Sir > > Mistakes are not improbable, however I can not condone your > inadequate backup procedures. Its totally out of imagination that > you may have rouinned the RIPE database. You have to keep yourself > more dedicated to your work. This cannot stand as is. It is important to understand that the data in the RIPE database is maintained by the users of the database, and not by the RIPE NCC itself. As was already said by the RIPE database administrator: > We would like to stress that the RIPE NCC operates the RIPE Database, > however the responsibility for managing user data lies with the users. I suspect that the problem is that a backup cannot easily be used to restore the current "desired" state (minus the unfortunate update). The reason is that old NIC handles may have been recycled, i.e. reassigned to another object, and that object can subsequently have been referenced by an explicit use of that recycled NIC handle in an update. Sure, the subsequent updates could perhaps be doctored to remedy this, but that is perhaps not easily or quickly done, and would also mean "intervention in the data maintenance" by the RIPE NCC. I would thus tend to turn the argument above on its head and say that we collectively, the maintainers of the data in the RIPE database, have made inadequate use of the already available methods for protecting the data we maintain there from unauthorized updates or removals. Regards, - Håvard

4 4

Events of Tuesday, 27 October 1998.
by RIPE Database Administration 02 Nov '98

02 Nov '98

Dear Colleagues, On Tuesday, 27 October 1998, a user of the RIPE Network Management Database accidentally deleted many [person] objects. The objects affected were either not maintained by a 'mntner' object or were maintained by a 'mntner' object with an authentication mechanism of NONE. The RIPE NCC has investigated this and learned that it was a genuine mistake by this user. When a person object is deleted, its nic-handle is available for re-use immediately. Therefore, you may find that a nic-handle now points to a different person or role object. We recommend that you check the consistency of your data i.e. see if the nic-handles in the admin-c, tech-c, and zone-c lines of your objects refer to the correct person or role objects. We understand that an attempt was made to re-create the deleted objects by the individual who originally deleted them. Please search for your person objects using both the name and the nic-handle. It is possible that your object has been re-created, but with a different nic-handle. If so, you must update your inetnum, route, mntner, etc. objects that refer to the previous nic-handles in the admin-c, tech-c, and zone-c lines. You should replace the old nic-handles with the new ones. To check whether a certain nic-handle is referenced anywhere, you can use the inverse look-up flag. For example, you can do: "whois -i admin-c,tech-c,zone-c <nic-handle>" to see if that nic-handle is referenced anywhere as admin-c, tech-c, or zone-c (e.g. as in domain objects). If one or more of your person and/or role objects have not been re-created, we suggest that you first try to re-create them using the original nic-handle. This way, you do not have to change any of the references that use this nic-handle. If the nic-handle has been re-used, you should try to create the person or role object with the AUTO-1 keyword. In this case, however, you will have to change the inetnum, domain, or other objects that reference this nic-handle to point to the new one. (Use the the "-i" flag mentioned above to find which objects reference the old nic-handle). The RIPE NCC shall, as a part of its service to the RIPE community, support users to restore their data to a consistent state. We estimate that this should take about one working day. Please note that it is the users responsibility to restore their own data. We would like to stress that the RIPE NCC operates the RIPE Database, however the responsibility for managing user data lies with the users. Mechanisms to protect data against unauthorised or unintentional modification exist for several years now. The RIPE NCC has always emphasized the importance of authentication mechanisms and has been recently working to introduce even stronger procedures. We encourage users to deploy these authentication mechanisms. Please read ripe-157, section 2.3 on how this works. If you have any problems with the RIPE Database, please send an email to <ripe-dbm(a)ripe.net>. Regards, The RIPE NCC Database Group.

3 2

RE: Database performance - an update.
by Mike Norris 30 Oct '98

30 Oct '98

> I shall begin by saying that I wrote the two messages to > the local-ir, lir-wg and db-wg yesterday. I wrote them > for, and on behalf of, the RIPE NCC Database Group. I > realise that many of you are unhappy that these messages > had no personal signature. The responsibility for this > oversight is mine. In all previous postings to these lists, > I have signed my name. I shall continue to do this. Ambrose, you may not have put your name at the end of the mail, but we knew it was you anyway - your style is inimitable and your sense of humour unique. Keep up the good work and watch out for ghosts this weekend. Cheers. Mike Norris

1 0

Database performance - an update.
by RIPE Database Administration 30 Oct '98

30 Oct '98

Dear Colleagues, I shall begin by saying that I wrote the two messages to the local-ir, lir-wg and db-wg yesterday. I wrote them for, and on behalf of, the RIPE NCC Database Group. I realise that many of you are unhappy that these messages had no personal signature. The responsibility for this oversight is mine. In all previous postings to these lists, I have signed my name. I shall continue to do this. You already know what happened on Tuesday, 27 October, last. Yesterday morning, we had no option but to do preventative maintenance of the database; this was not related to the activity of any user last Tuesday. However, nearly all updates were stopped until around 1630 MET. Unfortunately, there was a higher than usual number of updates sent to the database yesterday. These two facts had one result: a very long delay before your updates were processed. At this time there is still a long queue of unprocessed updates, but updates are being processed again. Obviously, a faster, more powerful machine would process updates quicker than the current machine. On Tuesday, the problem was that it was possible to delete objects referenced somewhere in the database before deleting all references to that object. Imposing such a constraint (referential integrity check) has been already been considered. Tuesday's incident was not a security problem - anyone with the correct authorization can still make a mistake. At the previous RIPE meeting in Edinburgh, we announced new software, including a referential integrity check, and new hardware, a Sun E450 machine. We intended to deploy both of these at the end of October or early November. We had bad luck that several unrelated problems occurred this week, just before the planned deployment of the new code and the new machine. Now, a status report. All your updates have been queued - there is no need to re-send them. We estimate that the situation should be back to normal tomorrow, if not sooner. We still plan to introduce the new software and hardware next week. We thank you all for your support in the past and we assure you of our continuing efforts to support you in your work. Your sincerely, Ambrose Magee -------------------------------------------------- for, and on behalf of, the RIPE NCC Database Group.

1 0

Objects affected by events of 27/10/1998.
by RIPE Database Administration 30 Oct '98

30 Oct '98

Dear Colleagues, We keep detailed logfiles of all updates to the RIPE Database. We processed these files to produce lists of all objects affected by the incidents earlier this week. We have decided to make these files public, to support you when ensuring that your data is consistent. These files contain only the name and nic-handle of affected person objects. These files may be obtained by anonymous ftp from: ftp://ftp.ripe.net/ripe/dbase/981027-incident.Delete ftp://ftp.ripe.net/ripe/dbase/981027-incident.New ftp://ftp.ripe.net/ripe/dbase/981027-incident.Update There are also gzip'ed versions available, using the same filenames, with the .gz extension. If you have any questions, please contact <ripe-dbm(a)ripe.net>. Regards, Ambrose Magee ----------------------- RIPE NCC Database Group

1 0

Big delays with <auto-dbm>.
by RIPE Database Administration 29 Oct '98

29 Oct '98

Dear Colleagues, Last night, there was a higher than usual number of messages to <auto-dbm(a)ripe.net>. Thus, there were some delays in processing these messages. This morning, we carried out some unrelated preventative maintenance and this has led to more delays. Your messages are being stored and will be processed as soon as possible. BTW, this maintenance was not linked to the recent actions of any user of the RIPE Database :-) If you have any questions, please contact <ripe-dbm>. It is either Murphy's Law, or Hallowe'en, or the Great Pumpkin, or whatever, that this happened this week. We thank you for your patience. Regards, --------------------------- The RIPE NCC Database Group.

1 0