Re: What date format for d/u/metadata ? Re: "Entry: NA" in debian/upstream/metadata
Hi Steffen,
On 2021-03-05 21:19, Steffen Möller wrote:
> Am 05.03.21 um 16:13 schrieb Andrius Merkys:
>> On 2021-03-04 23:12, Steffen Möller wrote:
>>> Somewhere else was the suggestion made to also add a time stamp. This
>>> makes perfect sense for the NA/~ and in that case, if that date was
>>> specified, we know that unknown is a confirmed unknown. For entries that
>>> are found, we should possibly just rely on git blame in salsa.
>> Exactly. This was my point. Because if someone stumbles upon a timestamp
>> from 3+ years ago, one may check the registry to see if the entry is
>> still not there. If the entry is still missing, one would update the
>> timestamp to let everyone else know "hey, I have checked it, and it is
>> not there". Otherwise one's effort will be lost, and the next one who
>> sees a missing entry may repeatedly drain one's time looking.
>
> Since I was just active on pigx-rnaseq for the thread on guix, I came up
> with
>
> Registry:
> - Name: OMICtools
> Entry: OMICS_33677
> - Name: conda:bioconda
> Entry: NA
> Checked: Fri, 05 Mar 2021 20:06:08 +0100
> - Name: guix
> Entry: pigx-rnaseq
> - Name: bio.tools
> Entry: NA
> Checked: Fri, 05 Mar 2021 20:07:04 +0100
>
> But, donno, this RFC 5322 is barely parseable by eye, even though this
> is how we typically put dates in Debian (you get this via 'date -R').
> Much more readable though would be `date --rfc-3339=date`
I would also vote for RFC 3339. RFC 5322 admittedly removes some
ambiguity (as confusing YYYY-MM-DD for YYYY-DD-MM), but is not so easy
to read/write. RFC 3339 is also widely used in Debian, for example, for
appending timestamps to source package versions and package diff files [1].
> Registry:
> - Name: OMICtools
> Entry: OMICS_33677
> - Name: conda:bioconda
> Entry: NA
> Checked: 2021-03-05
> - Name: guix
> Entry: pigx-rnaseq
> - Name: bio.tools
> Entry: NA
> Checked: 2021-03-05
>
> but do our American friends understand that this is not May? And we do
> not need the time, as in
>
> 2021-03-05 20:14:12+01:00
>
> I would start without the time and then add it if needed - but as I
> said, the art is to eliminate the NAs in the respective
> registry/repository and for that, the time of the day does not really
> matter, I tend to think.
Dates without time have a total of 48 hours of uncertainty due to time
zones (if my calculations are correct). Most likely this uncertainty
could be ignored for this particular application.
> A pending question is if we need a "<rejected>" as in "This entry is not
> going to be added to that repository". I personally do think so and
> consider this information more important than the NA since a repeated
> request likely annoys someone on the other end.
Some messages ago [2] I have suggested introducing "Status" field for
indicating special states of entries, such as not found, rejected,
pending and like. Such field would completely remove the need to place
non-ID information in "Entry" field. What do you think about it?
[1] http://ftp.us.debian.org/debian/dists/unstable/main/source/Sources.diff/
[2] https://lists.debian.org/debian-med/2021/03/msg00035.html
Best,
Andrius
Reply to: