IQSS logo

IRC log for #dataverse, 2020-11-10

Connect via chat.dataverse.org to discuss Dataverse (dataverse.org, an open source web application for sharing, citing, analyzing, and preserving research data) with users and developers.

| Channels | #dataverse index | Today | | Search | Google Search | Plain-Text | plain, newest first | summary

All times shown according to UTC.

Time S Nick Message
04:22 foobarbecue joined #dataverse
08:44 juancorr joined #dataverse
12:08 donsizemore joined #dataverse
12:12 donsizemore joined #dataverse
12:14 donsizemore @pdurbin @pameyer I'm good with whatever you all want WRT the PR. It started out as a simple change in the installation docs to recommend RHEL/CentOS8 at Phil's request =)
13:48 juancorr joined #dataverse
15:33 pameyer joined #dataverse
15:34 pameyer @donsizemore - a "simple change" can be deceptively non-simple sometimes
16:02 poikilotherm Oi pameyer. Q?
16:02 pameyer sure
16:02 pameyer ... useful answers not guaranteed though
16:17 pdurbin joined #dataverse
16:18 pdurbin pameyer donsizemore: I moved the CentOS 8 pull request to QA this morning. Thanks, Don, for merging my changes. I think we're good. Mostly just recommending CentOS 8 and upgrading docker-aio to it along the way.
16:19 donsizemore One day we'll move to Debian 10
16:19 pdurbin oh, merged already. nice
16:20 pameyer that's quick
16:36 poikilotherm pameyer you are handling images of neuroscience, aren't you?
16:38 pameyer poikilotherm - nope, mainly x-ray diffraction images
17:03 pdurbin poikilotherm: your users have some of these images?
17:03 poikilotherm Oh that might be related enough. DICOM?
17:07 pameyer nope - adsc, mar, cbf, eiger
17:08 pameyer there was a push for hdf5 standardization a while back, but some folks were unclear about what hdf5 was (aka - more analgous to xml than jpeg)
17:18 poikilotherm Ok. We had someone unrelaxed about RDM as a concept of making data FAIR. But on the other hand doing RDM all the time with very interesting tools like XNAT, Datalad and DICOM.
17:19 poikilotherm Storing lots of patient data (mri, ...) that cannot be shared easily.
17:21 poikilotherm So I thought you guys might have similar stuff to handle. Similar scientific culture and problems, calming down people's worries about more work...
17:22 pameyer for patient data, that's probably closer to trsa than anything I've worked on
17:23 poikilotherm TRSA was UNC/Odum/donsizemore, right?
17:23 pameyer other than rants about ethics, de-identification, differential privacy and related topics
17:24 pameyer yup - I'm pretty sure there's active development on TRSA.  there was some overlap for storage locations (we were concerned about replication of public data too)
17:36 donsizemore @poikilotherm TRSA is Odum/akiosone =) he did a demo to IQSS this morning
17:38 donsizemore @poikilotherm https://github.com/OdumInstitute/trsa-web/blob/master/ImpactTRSAConcept.pdf
17:52 poikilotherm Donsizemore: sounds great. Has there been a recording of this demo?
17:52 donsizemore not to my knowledge
17:52 poikilotherm I know the concept paper, but I'm not sure that's what scientist are likely to understand :-D
18:11 poikilotherm Ok other topic. pdurbin, Q?
18:13 poikilotherm I know it's tech hours soon...
18:30 pdurbin Tech hours is 90 minutes away but in 30 minutes I need to present some thoughts on pain points of JSF. Any ideas?
18:40 pameyer templating, performance/caching for frontend bits, doctype confusion, random bits of javascript/css, static files with session cookies that make caching problematic
18:41 pameyer off the top of my head, and with the disclaimer that I'm not a java developer
18:42 pameyer static files from jsf included components having cookies is probably the one with the most ops impact
18:46 poikilotherm Not sure JSF is that bad or if it is just the way we use it. Not very experienced on that field. What's the idea behind it? Switch to REST+Webcomponents?
18:47 poikilotherm Pdurbin I'll stash my question for later use... ;-)
18:49 pameyer good point - I'm not knowledgable enough about JSF enough to know how much of that to attribute to it, vs other things
18:49 pameyer cookies for static files was definately a pain point that stuck in my head a while back when looking at frontend optimizations though
18:58 pdurbin I didn't know about the cookies for static files thing.
19:01 pameyer I took a look at Harvard Dataverse a few minutes back, looks to still be the case
19:02 pameyer but as poikilotherm pointed out, that may not be JSF's fault
19:02 pdurbin Yeah, I don't know.
19:03 pdurbin This still seems valid: https://github.com/IQSS/dataverse-helper-scripts/blob/e72145f1d611ec1fd9bd64cdd313494f6c013fda/system-thoughts/issues-with-primefaces.md
19:04 pameyer that's a 404 for me
19:05 pdurbin hmm, maybe try https://github.com/IQSS/dataverse-helper-scripts/blob/master/system-thoughts/issues-with-primefaces.md
19:07 pameyer that works
19:07 pameyer checkboxes...
19:08 pdurbin :)
19:08 pdurbin my JSF meeting got pushed back so I guess I have more time to think about it
19:08 pameyer some of those I'm not competent to judge (server side memory usage of tree, etc)
19:09 pameyer but most of them look still valid to me
19:16 poikilotherm Indeed, looks like all of this is still an issue. Although Angular dropped popularity and now there is Vue.js. Not to forget: webcomponents as w3c standard, so easier to get rid of all these frameworks that are another vendor lock in.
19:17 pameyer I haven't looked into webcomponents at all
19:18 pdurbin I use Web Components at https://github.com/IQSS/dataverse-tv such as the template tag in index.html.
19:23 pameyer pdurbin: are you familiar with how django and flask configure static files (so they can go out from apache/nginx directly, vs the app server)?
19:24 pdurbin vaguely
19:24 pameyer not to suggest rewriting everything to do that, but it's a different way of looking at them
19:25 pameyer template tag/variable for the URL root of static files, and some mechanics to configure/copy out of app directory
19:26 pdurbin no objection :)
19:27 pdurbin But I also have no objection to rewriting everything, if we ever have the time.
19:28 pameyer that's always the qualifier...
19:30 pdurbin I'm just happy that there's some chatter about it. :)
19:30 pdurbin well, about pain points at least
19:31 pameyer I don't know if anybody's kept up with the frontend performance comparison for dataverse vs other repositories
19:32 pameyer but that reminded me about not optimizing when you don't need to
19:33 pameyer and if it's kept constant, non-cachable static files are likely not worth worrying about
19:34 pdurbin Yeah, I'm thinking more of developer productivity, a strong ecosystem, testing, etc.
19:34 pdurbin But sure, performance should always be considered. "Speed is a feature" to quote the creator of Stack Overflow.
19:35 pameyer keeping complexity down usually helps both
19:36 pameyer it's not an easy problem though
19:41 pdurbin "We decided to replace JSF with a Web Components/Polymer-based application for initiating Web UI development cycles for different reasons:" https://www.nuxeo.com/blog/jsf-deprecated/
19:43 pameyer I'm not sure I agree with "client side for better user experience" ; but I think that ship sailed long ago
19:43 pdurbin :)
19:44 pdurbin My silly DataverseTV app is client side. So is the Dataverse map. But I hear ya.
19:46 poikilotherm Testability is an big pain point with the current JSF+Prime faces approach
19:47 poikilotherm The tooling ecosystem around testing is relying on the industry standard use cases like Angular, Vue, plain HTML etc.
19:48 pameyer the main testability thing I remember being an issue was lack of useful element ids (at least, frontend testability)
19:48 poikilotherm And with the flexibility of metadata there might arise a need to have more flexibility than Primefaces offers in terms of how you can adapt the interface.
19:48 pameyer sometimes could be worked around with placeholder text, sometimes (checkboxes) not
19:49 pdurbin I agree but Slava recently released some Selenium scripts at https://github.com/GlobalDataverseCommunityConsortium/dataverse-test-suite
19:50 poikilotherm At some point in the future it might be a necessity to get Dataverse more modular. That might become easier when JSF stays out of the game. Decoupling front and backend.
19:50 pameyer interesting - I hadn't seen Slava's stuff
19:51 pameyer looks like he's got the same element id issue
19:53 pameyer although those also don't look like selenium I've written (js vs python), so I might be mis-reading
19:57 pdurbin tech hours in 3 minutes, everyone's invited
20:13 pdurbin This is interesting, using JSF to call over REST to the back end: https://www.youtube.com/watch?v=MPP34MIww2k https://www.slideshare.net/javajuneau/utilizing-jsf-front-ends-with-microservices
21:58 pdurbin Ok, folks, I'm out and it's a holiday tomorrow. See you Thursday.
21:58 pdurbin left #dataverse
23:00 nightowl313 joined #dataverse
23:06 pameyer pdurbin: thanks for the holiday reminder

| Channels | #dataverse index | Today | | Search | Google Search | Plain-Text | plain, newest first | summary

Connect via chat.dataverse.org to discuss Dataverse (dataverse.org, an open source web application for sharing, citing, analyzing, and preserving research data) with users and developers.