Time
S
Nick
Message
07:06 Virgile joined #dataverse
09:29 juancorr joined #dataverse
09:37 Virgile joined #dataverse
09:38 Virgile joined #dataverse
12:52 juancorr joined #dataverse
14:22 pdurbin joined #dataverse
15:36 sivoais joined #dataverse
16:21 poikilotherm
Good morning pdurbin Sir
16:21 poikilotherm
Had a chat with Slava earlier today
16:22 poikilotherm
Talked about coming up with some ideas to ignite a discussion about a community backed testing strategy
16:23 poikilotherm
Add some counter spells to avoid rotting code being untested as resources are limited at IQSS etc
16:24 poikilotherm
Making people responsible for their (larger) contributions
17:15 pdurbin
Spells against rot. I like it.
18:46 poikilotherm
pdurbin do you want me to test if I can switch #7695 to sth usable with 5.2020.6 and still have reasonable production defaults in microprofile-config.properties?
18:48 pdurbin
You mean take out :-1 or whatever?
18:50 poikilotherm
Aye
18:51 poikilotherm
It should be possible to put these into another source. And defaults go to that file usually
18:51 poikilotherm
I'd test that and not argue with scolapasta. Not worth it.
18:51 pdurbin
Right. We already have a place for defaults.
18:52 poikilotherm
Also I strongly believe that switching your application server version should not be such a pain in the ***
18:52 poikilotherm
Maybe we need to make things easier somehow
18:53 pdurbin
yes, please
18:54 pdurbin
Shouldn't it be java -jar payara.jar?
18:54 poikilotherm
🤩
18:55 poikilotherm
Java -jar dataverse.jar
18:55 pdurbin
there you go :)
18:57 poikilotherm
Actually it shouldn't be that hard to  do with Payara embedded
18:57 poikilotherm
https://blog.payara.fish/using-payara-embedded-and-payara-micro-with-maven
18:58 poikilotherm
Maybe a point for the testing strategy
19:07 pdurbin
Adam Bien talks about a t-shirt from the old days that said java -jar glassfish.jar
19:07 poikilotherm
Btw not merging the PR will result in a little fork for me. I have to rely on the updated Payara version for the container images, as it has my rewritten directory config onboard.
19:07 pdurbin
Ah, you might want to mention that in the PR.
19:07 pdurbin
To me it shows there's a reason to upgrade. You're doing it for a reason. Not just because.
19:08 poikilotherm
Yeah. However.
19:09 poikilotherm
I'll need a little fork or at least special branch anyway, as I want to use GitHub Actions for the image building, security scans etc
19:09 poikilotherm
Does hurt to update the Payara version in there, too
19:10 poikilotherm
Doesn't
19:10 poikilotherm
Autocorrect
19:30 pdurbin
Do you wait for features you want or always update to the latest?
19:38 poikilotherm
Features and sec updates.
19:38 poikilotherm
And of course it shouldn't get too dusted
19:39 pdurbin
You must have been going crazy with Glassfish 4.1.
19:39 poikilotherm
It worked.
19:40 poikilotherm
But we can transform now to a modern stack. Use all the goodies of the recent Jakarta Ecosystem
19:40 pdurbin
Please keep dragging us into the modern era.
19:41 poikilotherm
Actually I think it would be a good thing to run e.g. demo.dataverse.org on a top recent stack
19:42 pdurbin
I keep saying I'd love for it to be powered by dataverse-kubernetes.
19:42 poikilotherm
As each Payara is updating deps that might introduce bugs etc, we can fail early
19:42 poikilotherm
Plus tests tests test
19:42 pdurbin
Oh, well for that we should use beta.dataverse.org (which should also run the develop branch)
19:43 poikilotherm
What happens when a sec problem is discovered in an old version of Payara and we need to switch fast, but haven't tested before
19:43 pdurbin
The demo server needs to be relatively stable for training, etc.
19:43 poikilotherm
I humbly disagree.
19:43 poikilotherm
You need traffic and user interaction to learn what makes things break
19:44 poikilotherm
Major flaws may never leave dev
19:44 pdurbin
In that case, I'll send our trainers after you when demo breaks during a session with an important stakeholder. :)
19:44 poikilotherm
It's the tricky parts, the dark areas where there might not yet be test etc
19:45 poikilotherm
I'm not trying to say let's fire up latest dev there
19:46 poikilotherm
Use a stable version of data verse and a stable version of the dependencies like Postgres, Solr and Payara after extensive testing
19:46 poikilotherm
(More testing than we do now...)
19:47 poikilotherm
Make it as stable as possible by any means. If it breaks, that would be sth you don't want to see in production
19:47 pdurbin
And maybe do something with logs. I'm thinking of rollbar or sentry.
19:48 poikilotherm
Maybe. The logging on demo might be necessary to tune for more verbosity, making errors traceable.
19:49 poikilotherm
Ah, visions. Don't listen to me, I'm dreaming to many big things.
19:51 pdurbin
Did you want to say more about the testing ideas? You and Slava talked?
20:13 poikilotherm
Yeah
20:14 poikilotherm
We talked
20:14 poikilotherm
Do you want to hear more?
20:14 pdurbin
I'm thinking. :)
20:14 pdurbin
I mean, at some point, yes.
20:14 poikilotherm
ðŸ¤
20:14 pdurbin
We could reach more people during a community call.
20:15 pdurbin
Or hell, the community meeting will be here before we know it.
20:15 poikilotherm
An early idea was to create a Google Doc first and write some ideas down...
20:15 poikilotherm
Some thoughts and mindsets
20:16 poikilotherm
To summarize why this came up: more and more features are coming in that IQSS didn't ask a for, does not get paid for and maybe has no benefit from.
20:17 poikilotherm
We should not rely on IQSS do all the hard maintenance work for free
20:17 poikilotherm
When the community wants certain features, so be it, but then the community needs to ensure maintenance and sustainability
20:18 pdurbin
A noble vision.
20:18 pdurbin
I do think that tests help ensure maintenance, if that's where you're going with this.
20:19 pdurbin
We could probably break OIDC at not know it.
20:19 poikilotherm
Aye. That's the only way to go IMHO, as usually paying IQSS for stuff won't be an option
20:20 pdurbin
and*
20:20 pdurbin
Or Handles.
20:21 poikilotherm
Yes
20:21 pdurbin
Or a whole slew of feature we don't use in production.
20:21 poikilotherm
So many code parts
20:21 pdurbin
Swift. The list goes on.
20:22 pdurbin
It's not even a code thing. Sometimes the community has questions about Handles. Who knows how it works. :)
20:23 pdurbin
not even *just* a code thing, I mean
20:23 poikilotherm
Yeah. And if we don't want to go the modularity way of things to decouple the maintenance, which would mean a major refactoring, we need another strategy
20:23 pdurbin
Are there any community-contributed features that are well maintained and sustainable?
20:24 poikilotherm
Thin ice 😎
20:24 pdurbin
I feel like Handles have worked out pretty well.
20:25 poikilotherm
Honestly... I can't think of one from the top of my head
20:25 pdurbin
Might be too early to say for OIDC.
20:26 poikilotherm
Does well maintained mean there are good tests around?
20:26 poikilotherm
Then Handle and OIDC aren't well maintained 🙈
20:26 pdurbin
:)
20:26 poikilotherm
S3 works pretty good AFAIK but there aren't many tests...
20:26 pdurbin
I think it's more than that. Can the users get support on the feature?
20:27 pdurbin
If you open an issue, does it get fixed?
20:27 poikilotherm
Uhm. Just by the community and asking politely
20:27 poikilotherm
If it email Dataverse Support are you going to help me with OIDC?
20:27 poikilotherm
If I...
20:28 pdurbin
An attempt will probably be made. :)
20:28 poikilotherm
Oh that's very nice and polite!
20:28 pdurbin
best effort
20:29 pdurbin
In practice I'd probably ask you if you mind being on the ticket so you can reply.
20:29 pdurbin
The OpenShift code was so little maintained that I didn't feel bad removing it.
20:29 poikilotherm
Definitely.
20:30 pdurbin
If code isn't used, it should be removed.
20:31 poikilotherm
I would rather not include my container stuff in IQSS upstream
20:31 poikilotherm
But due to archictural decisions of the maven project, we can't go otherwise
20:31 pdurbin
Right. I mean, you might be able to use an overlay.
20:32 poikilotherm
Overlay?
20:34 pdurbin
I *think* this is an example: https://github.com/sbgrid/sbgrid-dataverse/blob/empty-overlay/mod-sbgrid/pom.xml
20:35 pdurbin
Basically, you overwrite certain files in your local fork.
20:35 poikilotherm
https://maven.apache.org/plugins/maven-war-plugin/overlays.html
20:36 poikilotherm
I'm not sure this is possible with the stuff I'm up to. Maybe worth a try.
20:36 poikilotherm
Still it's a fork
20:36 poikilotherm
And you don't release to Maven Central
20:37 pdurbin
Yeah, still a fork. And harder for you to maintain. It probably makes sense to keep getting upstream to accept your changes.
20:37 poikilotherm
And I don't know how easy it would be done for arbitrary branchrs
20:37 pdurbin
No idea. I just know the SBGrid guys had it working with Dataverse. Seemed like a nice option.
20:38 poikilotherm
At least it's not a hard fork
20:39 poikilotherm
It might be worth a shot for my stuff
20:40 poikilotherm
Thx for pointing me to this. I only knew about parent poms
20:41 pdurbin
Sure. We could also get pameyer's take on it. I think he's still using it.
20:42 poikilotherm
It may cause headaches how to use this with dev branches of data verse.
20:42 poikilotherm
There is no parent pom triggering the build of the dataverse module
20:43 pdurbin
We used to have a parent pom in DVN 3.
20:44 poikilotherm
I wonder what happens if we add a parent bom to upstream, I could reuse in dv-k8s
20:44 poikilotherm
It should still be possible to build the project on its own
20:44 pdurbin
https://github.com/IQSS/dvn/tree/develop/DVN-root has a pom.xml and so does DVN-web, which is where all the code was.
20:57 pdurbin
Anyhoo, time for me to head out. See you all tomorrow.
20:57 pdurbin left #dataverse
21:02 poikilotherm
Cu
