Implementation plan for replacing "changed:" with "last-modified:" and "created:" published
Dear colleagues, The RIPE Database Working Group requested the RIPE NCC to replace the “changed:” attribute in the RIPE Database with “created:” and “last-modified:”. Following discussions at RIPE 69, the RIPE NCC has met with the working group chairs and finalised an implementation plan consisting of three phases: 1) introduce the new attributes “created:” and “last-modified:” 2) make the "changed:" attribute optional 3) deprecate the "changed:" attribute. The time lines and documentation for each phase are detailed here: https://labs.ripe.net/Members/tim/deprecating-the-changed-attribute-in-the-r... We look forward to any questions or comments you might have. Kind regards, Tim Bruijnzeels Assistant Manager Software Engineering RIPE NCC
On Mon, Feb 02, 2015 at 10:37:41AM +0100, Tim Bruijnzeels wrote:
1) introduce the new attributes ?created:? and ?last-modified:? 2) make the "changed:" attribute optional 3) deprecate the "changed:" attribute.
good to see this move ahead, the automated time stamps will help reduce confusion.
The time lines and documentation for each phase are detailed here: https://labs.ripe.net/Members/tim/deprecating-the-changed-attribute-in-the-r...
We look forward to any questions or comments you might have.
What type of objects will be affected by the proposed change (assume: all)? What type of objects will be subject to the --list-versions optiuon? Both might have data protection aspects (cf. Piotr's mail <20140912093643.GH8166@hydra.ck.polsl.pl> dated 12 Sep 2014) and the labs article is silent about this. Minor issue: the removal of the "changed:" attribute in phase 3 or any otehr change in the previous phases will not be changing the time stamps, following the spirit "the RIPE NCC will make sure that this attribute is not modified in case of bulk changes that have no semantic meaning"? -Peter
Dear working group, On 02 Feb 2015, at 11:23, Peter Koch <pk@DENIC.DE> wrote:
On Mon, Feb 02, 2015 at 10:37:41AM +0100, Tim Bruijnzeels wrote:
1) introduce the new attributes ?created:? and ?last-modified:? 2) make the "changed:" attribute optional 3) deprecate the "changed:" attribute.
good to see this move ahead, the automated time stamps will help reduce confusion.
The time lines and documentation for each phase are detailed here: https://labs.ripe.net/Members/tim/deprecating-the-changed-attribute-in-the-r...
We look forward to any questions or comments you might have.
What type of objects will be affected by the proposed change (assume: all)?
yes, all
What type of objects will be subject to the --list-versions optiuon?
Both might have data protection aspects (cf. Piotr's mail <20140912093643.GH8166@hydra.ck.polsl.pl> dated 12 Sep 2014) and the labs article is silent about this.
--list-versions and --show-version are not available for Person and Role objects for this reason, and we have no current plans to change that. That said, technically, the working group could explore the idea of authenticated queries. In theory we could allow authorised maintainers of objects to see the history of these objects. I believe that data protection concerns would not apply in that case. We may also be able to supply more information in future for *authenticated* queries only - such as which maintainer, or when auth is introduced on person objects, which person made a change. We currently do get questions about updates from users from time to time, because it's not clear to them which of their colleagues made a change.
Minor issue: the removal of the "changed:" attribute in phase 3 or any otehr change in the previous phases will not be changing the time stamps, following the spirit "the RIPE NCC will make sure that this attribute is not modified in case of bulk changes that have no semantic meaning"?
Yes, indeed. In earlier discussions we concluded that if an attribute is removed altogether, then by definition that attribute is no longer meaningful. Kind regards, Tim Bruijnzeels
On Mon, Feb 02, 2015 at 12:02:28PM +0100, Tim Bruijnzeels wrote:
Both might have data protection aspects (cf. Piotr's mail <20140912093643.GH8166@hydra.ck.polsl.pl> dated 12 Sep 2014) and the labs article is silent about this.
--list-versions and --show-version are not available for Person and Role objects for this reason, and we have no current plans to change that.
thanks, Tim, for clarifying this. For person/role objects, there remains a tiny aspect then that will be slightly different from before. While many maintainers keep all changed: attributes and often enough the first one equals the creation date, an explicit "created:" (modulo all uncertainties about legacy data) will show a person's "age" in the DB. That's not necessarily a showstopper (could be mitigated if need be), but could benefit from a sentence or two. There are lots of reasons why one might _want_ to see this, but the real need isn't as obvious to me as is the last-modified: timestamp. And, btw, that also holds for the output of person handles in the context of other object types' histories.
That said, technically, the working group could explore the idea of authenticated queries. In theory we could allow authorised maintainers of objects to see the history of these objects. I believe that data protection concerns would not apply in that case. We may also be able to supply more information in future for *authenticated* queries only - such as which maintainer, or when auth is introduced on person objects, which person made a change. We currently do get questions about updates from users from time to time, because it's not clear to them which of their colleagues made a change.
Sounds reasonable to me. Don't know how many person objects are multi maintained, but for a 1:1 relationship, the scheme above should work well. -Peter
participants (2)
-
Peter Koch
-
Tim Bruijnzeels