Time
S
Nick
Message
01:40 andrewSC joined #dataverse
01:44 jri joined #dataverse
07:22 jri joined #dataverse
08:22 jri_ joined #dataverse
08:23 jri joined #dataverse
08:59 poikilotherm joined #dataverse
11:36 donsizemore joined #dataverse
11:38 donsizemore
@pdurbin i've decided it's better to keep the suggested branch and repo in the script, allow them to be overridden with runtime flags, and store only the datacite stuff in the env file. seems like the easiest workflow for developers
11:38 donsizemore
@pdurbin so lemme work on that (assuming the building doesn't wash away today)
11:48 pdurbin
donsizemore: ok, that sounds good. I think the most common use case around here is to pass in the branch with -b. Pull requests from outside IQSS are getting more common, though, so continuing to support -r to pass in the repo url will be nice, I think.
11:48 pdurbin
Plus, that way we don't have to mess with the docs as much. :)
11:48 pdurbin
Though we should explain in the docs the new "env" file.
11:48 pdurbin
If you need a hand with that, I'm happy to add it.
11:53 pdurbin
donsizemore: did you see that Odum is mentioned in the "Open Science Infrastructure WG 10/24 Call Agenda" doc? Please see https://docs.google.com/document/d/1bOVWBfhOiKGU2dYoHN_Pkpv5zPI6y819UISfkTqYpoQ/edit?usp=sharing
11:54 donsizemore
@pdurbin jon was asking if i was on the e-mail chain about that call. he's going to ask craig to add me to the fray
11:55 pdurbin
donsizemore: ah, to get the emails, you should click "subscribe" at https://github.com/whole-tale/whole-tale/issues/43 (or leave a comment)
11:56 donsizemore
done
11:57 pdurbin
cool
11:57 donsizemore
@pdurbin jon says there is also a mailing list
11:57 pdurbin
Is there? I'm not on it. Do they have public archives?
12:04 pdurbin
donsizemore: I just opened this about the mailing list: https://github.com/whole-tale/whole-tale/issues/46
12:17 cdsp-rmo
hey there
12:18 cdsp-rmo
pdurbin: do you have any news about the questions I had on https://github.com/IQSS/dataverse/issues/4593 ? (about multiple depositors)
12:19 cdsp-rmo
I fixed points 1 and 3 by adding "good" error messages about import
12:19 cdsp-rmo
but I'm still stuck with point 2
12:30 donsizemore joined #dataverse
12:41 pdurbin_m joined #dataverse
12:41 pdurbin_m
cdsp-rmo: you should go ahead and make a pull request
12:43 cdsp-rmo
okay
12:44 pdurbin_m
cdsp-rmo: thanks. Also, what is the status of https://github.com/IQSS/dataverse/issues/5090 ?
12:44 cdsp-rmo
I still had the log problem
12:45 cdsp-rmo
I had to comment out the login part
12:45 cdsp-rmo
gonna commit my changes, merge with latest develop, and check again if the problem persists
12:49 pdurbin_m
cdsp-rmo: perfect. Thank you! Julian and I agree that multiple depositors should be supported. Can you please open a new issue about this?
12:49 cdsp-rmo
okay
12:52 pdurbin_m
thanks!
12:53 cdsp-rmo
https://github.com/IQSS/dataverse/issues/5164
12:56 pdurbin_m
thanks! I'll add some details in a bit
13:01 jri joined #dataverse
13:03 donsizemore
@pdurbin_m i think i'm beating up on AWS too hard. i stuck in a sleep 10 to give the instance time to spin up, and i'm still occasionally getting timeouts
13:05 isullivan joined #dataverse
13:08 pdurbin_m
weird
13:09 pdurbin_m
Whoops, I forgot to kill my instance yesterday. Amazon bill. :/
13:11 poikilotherm
Mornin' guys :-) News from Europe, freshly backed from the videoconference with Slava ;-)
13:12 poikilotherm
I talked to Slava about the PID plugin they develop and this might a very interesting thing for #5093, too :-)
13:13 poikilotherm
They have a simple service implementation which is able to talk to the Dataverse API to collect the metadata, transform it to some kind of request a PID provider needs and put the resulting PID back into Dataverse
13:14 pdurbin_m
huh
13:14 pdurbin_m
is it open source?
13:14 poikilotherm
This of course needs some minor code changes in Dataverse (ringing the service, ...), but this sounds like a good approach to get all the custom PID stuff outside of the upstream project and away from maintenance hell for you guys.
13:15 poikilotherm
Yes, it is. Don't have a link yet, but Slava will send it to me tomorrow
13:16 poikilotherm
The benefit from this is that Dataverse only needs to talk over a stable API to a evolving service. This is easier for maintenance and especially for testing.
13:17 poikilotherm
And if I think about my plans on switching from DataCite MDS API to REST API, this would be a very welcome approach to have the functionality separated and guaranted not to blow up existing usage.
13:18 poikilotherm
This separate project is testable without a working dataverse installation, which makes it also much lighter to test
13:28 pdurbin
poikilotherm: sounds great. How would you like to track this exploration? Do you want to make a new issue and we'll put it in "Community Dev" and assign it to you?
13:29 poikilotherm
Let me take a look at the code first... ;-) And Slava said they wanted to put their changes to Dataverse into upstream, so they most certainly will open an issue or just reuse the existing one.
13:29 pdurbin
Ok. However you want to track it is fine. Thanks for working on this!
13:29 poikilotherm
Maybe I can encourage him to be more responsive on GH... ;-)
13:36 poikilotherm
Question from a colleague of mine: do you guys think it would be possible to get the slides or even a recording of https://hlrdm.library.harvard.edu/event/preparing-your-data-and-code-reproducible-publication-hands-workshop-researchers
13:36 pdurbin
Meh. He's flown over for two community meetings. He calls in a lot. He tours Europe giving talks about Dataverse. He's great.
13:37 poikilotherm
This could be very interesting for our own workshops we need to come up with for the time after day 0, the day our dataverse installation comes to life...
13:37 poikilotherm
Yes he is :-D
13:40 pdurbin
huh. "Co-sponsored with Harvard Dataverse"
13:40 poikilotherm
That's why I'm asking here... ;-)
13:45 pdurbin
poikilotherm: I think I learned about it from https://twitter.com/hlrdmp/status/1036982254436798465 ... Can you please reply to that tweet and ask about slides and recordings?
13:47 pdurbin
donsizemore: "Code Ocean Workshop - Preparing your data and code for reproducible publication: A hands-on workshop for researchers" ^^ ... seems to relate to the Whole Tale meeting we were talking about earlier.
13:48 pdurbin
isullivan: looks like you're the instructor for https://hlrdm.library.harvard.edu/event/open-science-framework-workshop-0
14:14 poikilotherm
pdurbin: Done. :-)
14:17 pameyer joined #dataverse
14:21 pdurbin
Huh. I don't see it.
14:21 pdurbin
Ah, here: https://twitter.com/poi_ki_lo_therm/status/1050389163860787200
14:26 pdurbin
poikilotherm: I just reached out to the person who tweeted this: https://twitter.com/jgolds2/status/1039927601479995396
14:38 jri joined #dataverse
14:39 pameyer
poikilotherm - any ideas why they didn't try to do api-compatability w\ ezid or mds for their PID stuff?
14:39 pameyer
ideas that can be publicly logged, I mean ;)
14:41 poikilotherm
actually these PID providers are pretty different in what they need and can do for you. it seemed not to be a good idea to be limited to the data that the ezid/mds api can transport
14:42 poikilotherm
da|ra might not be the best example, but e.g. ePIC allows you to have custom key-value-pairs directly attached to your metadata entry. this is not possible and not wanted by datacite
14:42 pameyer
thanks; makes sense
14:42 poikilotherm
and for a more concrete example
14:42 poikilotherm
i could not use the mds api to map to the rest api
14:42 poikilotherm
the mds api does not know about states
14:43 poikilotherm
so i would need to get that information from the metadata in dataverse anyway
14:43 jri_ joined #dataverse
14:44 poikilotherm
IMHO it really makes more sense to let people provide the mapping from the metadata in dataverse into the metadata schema from the pid provider themself and not fiddle around with possible limitations from the other apis
14:46 pdurbin
poikilotherm: will I be able to add a "no-op" PID provider that I can use in development? A PID provider that doesn't actually reach out over the network?
14:47 poikilotherm
why not?
14:47 pdurbin
cool
14:47 poikilotherm
what the provider thing is doing depends on what you develop for it. of course you could use a noop, which is simply skipping the communication with an upstream service
14:48 pameyer
"provide a mapping" - agreed; but I think that sorting out how to do this might be somewhat messy
14:48 poikilotherm
pameyer: sure. But that's the headache of the guy who needs to mapping and does not put load on you guys ;-)
14:48 poikilotherm
(which imho would be a bad idea)
14:49 poikilotherm
i really enjoy the idea of a separate web service, running independently
14:50 poikilotherm
this could even be used to provide custom made PIDs without any upstream support, just for internal use
14:50 poikilotherm
many usecases to think of
14:50 poikilotherm
but not bloating the upstream of dataverse :-)
14:51 poikilotherm
Alright guys, read you tomorrow :-)
14:51 pameyer
would need some changes on the dataverse side too (urls for different types of PIDs) - but less so than others
14:51 poikilotherm
Have a nice day
14:51 pdurbin
you too poikilotherm
15:30 jri joined #dataverse
15:51 cdsp-rmo
pdurbin: the pull request is finally there https://github.com/IQSS/dataverse/pull/5170
15:52 cdsp-rmo
gonna add some questions to it
15:58 pdurbin
cdsp-rmo: thanks! I just moved it over to code review in Waffle.
15:59 cdsp-rmo
wouhou
16:02 pdurbin
cdsp-rmo: since you're adding a new API endpoint, can you document it in the API Guide? The file is doc/sphinx-guides/source/api/native-api.rst
16:02 cdsp-rmo
oh no ...
16:02 cdsp-rmo
strange bug came again
16:02 jri joined #dataverse
16:02 cdsp-rmo
just deployed new war file, everything is fine
16:02 cdsp-rmo
stopped glassfish
16:02 cdsp-rmo
I wanted to do a new test, so I started glassfish
16:02 cdsp-rmo
can't log anymore
16:02 cdsp-rmo
><
16:03 pdurbin
:(
16:03 cdsp-rmo
it's like if stop then start glassfish breaks everything
16:04 cdsp-rmo
(ok for doc, that's why I left the checkbox)
16:04 pdurbin
It's no fun trying to get Glassfish to behave.
16:07 pdurbin
cdsp-rmo: thanks. Without any docs it's hard for QA to know what to test.
16:08 pameyer
I haven't gotten it to the point where I could try dataverse, but I was able to get a JSF hello world war to deploy on glassfish, payara and tomcat :(
16:09 pdurbin
pameyer: cool. Here's my non-JSF hello world that I got working fine on wildfly and payara: https://github.com/pdurbin/javaee-docker
16:15 pameyer
pdurbin: sounds like we're thinking of similar baby steps
16:15 cdsp-rmo
well, at least now I know what to change in the code to make it works :D
16:16 cdsp-rmo
so, even if I got the bug, I can still test my things
16:16 pdurbin
cdsp-rmo: sorry it's so painful
16:16 pdurbin
pameyer: I'd love to hear more about what you're thinking.
16:20 pameyer
pdurbin: partly inspired by the user (who's name I've forgotten) talking about ~"why have an install script when there are standard ways to deploy war files". figured the first step of that was to get war deployment working on a bunch of app servers
16:21 pameyer
didn't get farther than that; but next steps would've been seeing how much effort it would take to do 0_predeploy.sh , 1_postdeploy.sh in a generic way
16:21 pdurbin
pameyer: you might be thinking of https://github.com/IQSS/dataverse/issues/4523
16:21 pameyer
pdurbin - exactly
16:23 pameyer
also started wondering about pre/post deploy hooks in a war file (although that was more in the context of upgrades across versions). *looks* theoretically possible
16:24 pdurbin
It would be nice to have more open source models to look at. Jenkins is written in Java but I don't know if it's a jar or a war.
16:24 pameyer
war
16:24 pdurbin
Hmm. Maybe we can learn something from how Jenkins is deployed, then.
16:26 pameyer
works with tomcat - but I have a tendency to get mixed up about the distinctions between various types of java app servers
16:26 pameyer
but maybe
16:28 cdsp-rmo
well, it's time for me to go for sport
16:29 pdurbin
Because Dataverse uses Java EE APIs such as JMS a "full profile" Java application server is required. See "Full" vs "Web" profile at https://en.wikipedia.org/wiki/Java_Platform,_Enterprise_Edition
16:29 cdsp-rmo
doc is coming, I assure ! ;)
16:29 pdurbin
cdsp-rmo: bye! Thanks for the pull request!
16:29 cdsp-rmo
you're welcome
16:29 cdsp-rmo
bye
16:49 pdurbin
pameyer: not sure if that helps or not (full vs web). If you click the link to Adam Bien's repo from mine, I think he's on the right track with his "thin war" concept. In his world, the "full" app server is just another docker layer and the war file you're deploying is small. This all assumes Docker though.
16:53 pameyer
pdurbin: I think it does - but I probably need some time to digest it more
16:56 pdurbin
pameyer: basically the "fat jar" (or "fat war", I guess) model is used in Spring and deployed to Tomcat but Adam like the "thin war" model where you deploy to payara or wildfly.
16:56 pdurbin
likes*
17:09 donsizemore joined #dataverse
17:10 donsizemore
@pdurbin one more fix and i think we're good
17:12 pdurbin
donsizemore: yeah? Great!
18:24 jri joined #dataverse
18:26 jri joined #dataverse
19:14 jri joined #dataverse
20:24 jri joined #dataverse
21:13 jri joined #dataverse
21:25 jri joined #dataverse
21:41 donsizemore joined #dataverse
22:11 pameyer left #dataverse
