You're viewing an archived page. It is no longer being updated.
Thursday, 17 October, 14:00 - 15:30
Chairs: Denis Walker, William Sylvester
Scribe: Boris Duval
Status: Draft
The Chairs welcomed attendees to the session.
Edward Shrayne, RIPE NCC
The presentation is available at:
https://ripe79.ripe.net/archives/video/266
Rüdiger Volk, Deutsch Telecom asked about how to recognise default maintainers.
Edward Shrayne replied that it is recognisable as it is an additional mnt-by attribute sitting on the top-level resources.
Rüdiger Volk further commented that his company might need to hire temporary employees to help update those objects and that less experienced people might find it difficult to know which one is the default mnt. This might especially affect LIRs that have thousands of objects. He then asked the RIPE NCC to give an early notice to all LIRs about the clean-up to make sure they have enough time to prepare for it.
Edward Shrayne agreed that the workload will be consequent for some of the LIRs, but that the data needed to be cleaned as it has already been sitting there for four years.
Denis Walker, co-chair, asked Rüdiger why he was concerned about hiring external people to help clean the persons objects as they are currently locked.
Rüdiger Volk agreed that nothing could be done now, but that he had to be prepared to act fast when the objects will be unlocked as it will be a long and difficult job to complete.
Shane Kerr, NS1, pointed out that LIRs could anticipate the clean-up of locked person objects because they presumably have control over the objects that reference these. He added that they could dereference these objects and they will get cleaned up automatically.
Edward Shrayne added that what you can do right now to clean-up you own locked person objects is to create a new person role object, update your resource and once the person is unreferenced it will be cleaned up. However, he pointed out that it is a very slow process. He explained that when the RIPE NCC locked these objects three years ago, there was nearly a million of them. Now there is just over 600,000. Only 20 000 objects have been cleaned up in the last six months, since the last RIPE meeting. He concluded by saying that they could not wait for this process to happen on its own.
Cynthia Revström, Independent, asked if there was any solution to remove mnt-lower attributes that were not the default mtn?
Edward Shrayne answered that it is currently not part of the default mtn process, and that it is up to the LIR to organise the mnt-lower themselves. However, he mentioned that the RIPE NCC could look into it.
The presentation is available at:
https://ripe79.ripe.net/archives/video/269
Nurani Nimpuno, Asteroid, mentioned that she was pleased with the group forming the Task Force and wished them good luck.
Mahesh Aggarwal, RIPE NCC
The presentation is available at:
https://ripe79.ripe.net/archives/video/270
There were no questions.
Job Snijders, NTT
The presentation is available at:
https://ripe79.ripe.net/archives/video/272
Cynthia Revström thanked the presenter for raising this issue and proposing solutions to fix it.
Rüdiger Volk agreed that NWI 3 was not needed anymore. He added that his personal approach was “don't mess with other people's data”, and that he would prefer early notifications about potential issues over directly deleting user data.
Stavros Konstantaras, AMS-IX
The presentation is available at:
https://ripe79.ripe.net/archives/video/273
Rüdiger Volk commented that he was skeptical about throwing complex requirements (related to dynamic data) on to the RIPE database server as it could lead to system failure.
Stavros Konstantaras agreed that throwing complexity to the client side should be avoided whenever possible. However, he added that the RIPE Database does need to anything complex, as it already has information about the last modified field which keeps track of when an object is modified. The only thing that has to be done, Stavros argued, is to have a list of subscribers to notify relevant stakeholders when there is a change.
Job Snijders, NTT, thanked the presenter for tackling this issue and started a longer discussion with Stavros about the NRTM model and resolving.
They concluded that they could split NWI-9 in two parts: a legal part and a technical part.
Job Snijders added that the first task for the community would be to give everybody as close to real time access as possible for the dissemination of routing information. And the second task would be to look at updating NRTM 3, as the current version lacks authenticity and integrity, and need to be structured around modern day mechanisms.
Rüdiger Volk pointed out that putting complexity into the database server should be avoided and that until we have the technology to implement this properly. He added that the infrastructure should be kept simple and separated.
Stavros Konstantaras explained that this is why they proposed three approaches. One of these approaches, he added, is based on the publish-subscribe model used by Google. He further explained that this would help to take care of step one and free up time to focus on the development of NTRM 4.
William Sylvester, co-chair, thanked Stavros and proposed to continue the discussion on the mailing list.
Denis Walker, co-chair, declared consensus on NWI-10 and asked the RIPE NCC to implement it.
There were no comments.
There was no other business