Time
S
Nick
Message
08:38 poikilotherm joined #dataverse
09:06 jri joined #dataverse
11:32 pdurbin
poikilotherm: thanks for sending me your work plan.
11:32 poikilotherm
Morning pdurbin. Sure. You're welcome.
11:33 poikilotherm
Do you think it is reasonable?
11:35 pdurbin
Sure. It seems fine. Also, I'm glad to hear that you and Gustavo talked on the phone. I didn't know that.
11:36 pdurbin
How did that conversation go?
11:39 poikilotherm
That was last year... :-D It went pretty well and it brought up the idea of an internship ;-)
11:41 pdurbin
:)
13:47 donsizemore joined #dataverse
13:53 dataverse-user joined #dataverse
13:55 dataverse-user left #dataverse
14:09 isullivan joined #dataverse
14:29 pdurbin
Is anyone running OS X Mojave? I upgraded and now I can't start Glassfish 4.1 from Netbeans 8.2. I just opened this issue with details: https://github.com/IQSS/dataverse/issues/5431
14:49 drew-jhu joined #dataverse
15:02 Jim_ joined #dataverse
15:18 pameyer joined #dataverse
15:21 drew-jhu
happy new year, folks. @pdurbin - you mentioned a fake pid provider for dev environments & documented configuring for it here: http://guides.dataverse.org/en/4.10/developers/dev-environment.html#configure-your-development-environment-for-publishing . what is the effect of doing so? does it simply disable pid minting, or does it do something fancier?
15:21 pameyer
pdurbin: sounds like there's an issue with the credential file from `asadmin login` and netbeans
15:23 pameyer
@drew-jhu pdurbin's authorative on that one ;) but I'm pretty sure it just disables pid minting while acting like all the calls to the pid provider behave as expected
15:27 pdurbin
drew-jhu: the fake pid provider does as little as possible. As close to a no-op as possible. It doesn't reach out over the Internet. It's lazy. :)
15:30 donsizemore
@pdurbin so the fake pid provider is in fact a provided pid faker? it needs a good catty name
15:30 pdurbin
meow
15:30 drew-jhu
but it does have a different effect than `curl http://localhost:8080/api/admin/settings/:DoiProvider -X PUT -d JANKY` would...
15:30 donsizemore
@pdurbin the sean rowe virus comes to mind
15:31 drew-jhu
https://github.com/IQSS/dataverse/blob/19f19c25a95b03a6ba5b0d7c0cc4232f7cac77c2/src/main/java/edu/harvard/iq/dataverse/GlobalIdServiceBean.java#L81
15:31 drew-jhu
that's good to know
15:34 pdurbin
drew-jhu: you found the right line. JANKY won't work. Only FAKE.
15:34 pdurbin
... and the real DOI providers, of course.
15:35 pdurbin
(DataCite and EZID)
15:38 drew-jhu
so it would be accurate to say it "safely disables DOI registration" and is only available in versions >= 4.10, correct?
15:39 pdurbin
correct
15:39 pameyer
depends what you mean by "safely". if I'm remembering correctly, there's no visual indication to users that the PIDs are fake.
15:40 pdurbin
ah, good point
15:41 drew-jhu
i'm assuming that JANKY would eventually throw an error that FAKE wouldn't, so FAKE is safer in that sense
15:41 pdurbin
I haven't tried JANKY. :)
15:42 pdurbin
The fake PID provider is meant for developers. Please see "Do not require setup of PID credentials for development environments" https://github.com/IQSS/dataverse/issues/5327
15:44 pdurbin
Since moving developers from EZID to DataCite, we've had a lot of pain. EZID provided shared test credetials. DataCite does not. If you look at https://github.com/IQSS/dataverse/pull/5399/files you can see that I was able to replace a long paragraph with a single curl command.
15:57 drew-jhu
i think a bit more elaboration in the guide would be useful to make it clear that configuring :DoiProvider to FAKE (specifically) has the effect of safely disabling DOI registration, but that the UI does not currently provide any indication that the DOIs thus created are fake.
15:58 pdurbin
No objections. Would you like to create a pull request?
15:58 drew-jhu
was just about to offer.
15:59 pdurbin
Great! Please create an issue first. It's our way.
15:59 drew-jhu
wilco
15:59 pdurbin
:)
16:00 drew-jhu
i have to say, phil, you are a good community engagement pull-you-in person
16:02 pdurbin
one of us, one of us
16:03 pdurbin
I gave a somewhat recent talk on this stuff: http://wiki.greptilian.com/talks/2017/javaone-how-to-run-an-open-source-project/
16:43 drew-jhu
i'd say you walk that talk, phil. #thumbsup
18:02 pdurbin
:)
18:04 pdurbin
pameyer: the thing is, I'm not putting a password on glassfish. I never do.
18:07 pameyer
pdurbin: I probably got confused - somehow thinking new version of netbeans might be handling the glassfish interaction differently, and trying to do some kind of default login
18:07 pameyer
but yeah, that shouldn't matter for os changes
18:51 leonid
just to chime in the "fake" pid provider discussion. specifically re: "there's no visual indication to users that the PIDs are fake." I've said this before - they are not really "fake"; they are legitimate DOIs, issued in the name space (10.5072) that is specifically designated for assigning local, unregistered identifiers.
18:52 leonid
there is nothing more fake about this setup than, say, if you were using EZID or DataCite, but were using the same test name space. In both cases you end up with dois that are not registered anywhere.
18:53 leonid
the visual indication is the test name space. the analogy is the numeric notation for "localhost" - 127.0.0.1. It is a legitimate ip address; but you know that it's not registered in the DNS and only valid locally because it's purposefully reserved.
18:55 pdurbin
leonid: yeah, and as you've said, similar to private IP addresses: 192.168.x.x, 172.16.x.x, 10.x.x.x. These IP addresses are not fake but they are not to be used on the public Internet. Similar concept to 10.5072.
18:55 pdurbin
drew-jhu pameyer ^^
18:56 leonid
So, in retrospect, maybe naming it "test" instead of "fake" would have made more sense... I wish I thought about it back then.
18:56 pameyer
yeah; for devs I don't have any problems with that. but my expectation is that the number of users who'd recognize a test DOI namespace is even less than the number that would recognize a localhost ipv4 address
18:58 leonid
This is true. But, again, this is not unique to this new "fake" provider setup. Our default, out of the box setup for years was EZID, with a test account and this same test name space... same end result - ids that may (confusingly) look to some users as real, registered global ids.
18:58 leonid
(i recognize that this may be a problem;)
18:58 pameyer
very good point - this is not a new problem ;)
19:00 pameyer
I might be remembering wrong, but I think that test ezid dois would resolve (until they expired)
19:00 pameyer
don't think datacite test ones resolve, even if they're configured with a landing page
19:01 leonid
they are actually reserved, temporarily, yes. so, strictly speaking, i meant "the end long-term result is the same".
19:03 leonid
this means that using the test name space with EZID does have some value: you are testing your account credentials and/or the registration functionality in Dataverse.
19:06 pameyer
authenticating to external services is definately someplace I'd rather get it working before switching to production ;)
19:54 drew-jhu
not having an installation of 4.10 with which to test & develop a clearer understanding of how the new fake doi provider actually works is costing me here... i guess my thinking was that if i have configured my instance to use the fake provider, i'm probably not going to bother to also configure it to use the test namespace, in which case there would really be no visual indication that these DOIs won't behave like others do. TBH, this is no
19:57 pdurbin
not...
19:57 pdurbin
drew-jhu: I think you got cut off :)
19:58 drew-jhu
i just wanted to hit return after typing so many words and see if more talking was really necessary :)
20:00 pdurbin
TBH what though?
20:01 drew-jhu
one could provide guidance on whether it should be used on a staging environment or just production & whether it should be used in conjunction with a test namespace or if it obviates that, along with the pros & cons of mixing and matching these approaches, but that would probably be a blog post, rather than a quick note in the docs & would really end phil's victory of replacing a paragraph with a curl statement
20:02 pdurbin
heh
20:04 pameyer
maybe add something to the going live section of the docs? "don't use the FAKE provider in production"
20:05 pdurbin
I was feeling really bad for new contributors like Sheldon who were struggling: https://github.com/IQSS/dataverse/issues/5263
20:07 leonid joined #dataverse
20:08 drew-jhu
sheldon may be hitting up against the issue with the test environment DataCite just reported, but i'm not sure. once they told us about it, i stopped persuing that test
20:09 leonid
fwiw, the test name space ("authority
20:10 leonid
fwiw, the test name space (10.5072) is always set up by default, by the installer; by this line in setup-all.sh:
20:11 leonid
curl -X PUT -d 10.5072 "$SERVER/admin/settings/:Authority"
20:11 pdurbin
drew-jhu: just reported? Today? Where?
20:15 drew-jhu
day before yesterday, jon crabtree created test credentials for us under the GDCC account, but rteported he was unable to assign a prefix for it. yesterday, DataCite said, "We are aware of a bug that is affecting the assigning of prefixes to clients at the moment. We have a Github issue open and the technical team are working to resolve this as quickly as possible. I will let you know once it's fixed."
20:17 pdurbin
Huh. Did they link to the GitHub issue?
20:17 drew-jhu
keep in mind, i'm fairly new to dataverse myself & we only just migrated to datacite right before the holidays, so i'm not as familiar with how things are supposed to work as i'd like to be.
20:18 pdurbin
maybe it's this one: https://github.com/datacite/bracco/issues/126
20:19 drew-jhu
<shrugs>
20:21 pdurbin
90 repos under https://github.com/datacite so who knows
20:22 drew-jhu
...and most of them with codenames, to boot
20:30 drew-jhu
heh. earlier, i said, "one could provide guidance on whether [FAKE] should be used on a staging environment or just production" i meant "...or just development"
20:31 pameyer
gotcha :)
20:32 drew-jhu
hopefully we don't need to tell people not to use FAKE in production
20:33 pameyer
there have been a surprising number of times that folks have gone into production with the default robots.txt , and then had problems with getting their datasets indexed by google...
20:33 drew-jhu
good point.
20:35 pameyer
I'm glad you mentioned it though - because it would be a harder to fix failure mode than robots
20:37 drew-jhu
so, once i have a more fully functional test environment and a fuller picture of how thigs can work with the test datacite sandbox &/or the FAKE DOI provider, i would entertain the possibility of trying to draft more documentation or perhaps even a blog post, should that appear needful. in the meantime, if someone else is inclined to elborate on my elaboration in PR #5433, that's cool by me.
20:40 pdurbin
I'm fine with what you wrote and moved it to QA. I'm not sure if leonid or pameyer have looked at it yet. Here's a handy link: https://github.com/IQSS/dataverse/pull/5433
20:40 pameyer
pdurbin: I just took a quick look
20:40 pdurbin
drew-jhu: did you see the note about the config page in the issue?
20:40 pdurbin
"FAKE" appears there too. In the Installation Guide, I mean.
20:42 drew-jhu
i did & i think that's helpful. however, when i came across the curl line, i didn't immediately think to look elsewhere to see how it worked, so i do think that some elaboration at that point in the docs is useful
20:44 pdurbin
Sure. One thing I'll say about the "dev env" page is that we try to keep it short.
20:44 pdurbin
It used to be super long. I moved a lot of stuff into other pages recently.
20:44 pdurbin
We don't want to overwhelm new contributors with a super long page.
20:47 pdurbin
like we used to :)
20:48 pameyer
for what it's worth, I may have never read the original "dev env" stuff
20:49 drew-jhu
if you think terseness would be more valuable here, then that's cool too.
20:50 drew-jhu
it wasn't entirely clear to me a first glance whether FAKE was a magic word that would cause the code to behave differently or if it was just a brute-force way to keep it from butt-dialing anyone
20:56 drew-jhu
heh. it only now occurs to me that i did read that section & the part taken out is what alerted me to the possibility of getting test credentials for https://mds.test.datacite.org
20:58 pdurbin
Heh. Well, you're welcome to put whatever in your pull request. Want me to pull it out of QA so you can work on it some more?
21:01 drew-jhu
i'm disinclined to add more now unless someone else has a strong opinion about it. as i say, after i have more experience with the sandbox (and 4.10), i'd consider elaborating more *somewhere* (maybe not on that dev page, if you're trying to keep that short)
21:07 pdurbin
Sure, sounds good. Again, no objections to what you added so far.
21:11 poikilotherm joined #dataverse
21:51 pdurbin
folks, have a good weekend, I'm out
21:51 pdurbin left #dataverse
