Log of the #fcrepo channel on chat.freenode.net

Using timezone: Eastern Standard Time
* thomz joins03:16
* jjtuttle leaves06:15
* jjtuttle joins06:17
* jjtuttle leaves06:42
* jjtuttle joins06:44
* jjtuttle leaves07:50
* jjtuttle joins07:51
* jjtuttle leaves07:56
* jjtuttle joins08:00
* jjtuttle leaves08:01
* jjtuttle joins08:02
* jjtuttle leaves08:18
* jjtuttle joins08:22
* arebenji joins08:24
* dwilcox joins09:07
* whikloj joins09:08
* ajs6f joins09:10
* mikeAtUVa joins09:18
* tolloid joins09:23
* mikeAtUVa leaves09:29
* peichman joins09:35
* acoburn joins09:39
* acoburn leaves09:41
* acoburn joins
* mikeAtUVa joins09:44
* peichman leaves09:56
* peichman joins09:58
<ajs6f>acoburn: Are you still working with Apache Spark?10:08
<acoburn>ajs6f: yes
<ajs6f>acoburn: Have you seen doi:10.1093/logcom/exn044 ?10:09
acoburn: The relevance is that if inference is filling-in pullbacks in a category, that's a local calculation. That's something you can do with GAP.10:10
<acoburn>ajs6f: I haven't seen that
* tolloid leaves
<acoburn>ajs6f: that is a very interesting use case
<ajs6f>acoburn: GAS, I meant. Yes, the interest to me is a POV I've held for a long time— when you have the right formulation of a problem for work in a graph, you know it because all state becomes local.10:11
<acoburn>ajs6f: I've only just read the abstract, but I'm guessing the authors weren't using Spark (since the paper was from 2009)
<ajs6f>acoburn: Not only were they not using Spark, they have no interest in computation per se.10:12
<acoburn>ajs6f: are you thinking of the GraphX package that comes with Spark?10:13
<ajs6f>acoburn: Yes-ish. Things like this : https://spark.apache.org/docs/latest/graphx-programming-guide.html#pregel-api or in a different context, this: https://wiki.blazegraph.com/wiki/index.php/RDF_GAS_API10:15
<acoburn>ajs6f: the GAS approach is interesting. This is where I have all sorts of arguments with classic HPC folks about Spark/Hadoop vs. MPI10:20
<ajs6f>acoburn: What kind of arguments? It's two different things. That's like having an argument about hammers vs. screwdrivers.10:21
<acoburn>ajs6f: we have faculty who will _only_ use open_mpi; I keep telling them that for certain problems, that's great, but there are other tools out there10:22
* tolloid joins10:23
<ajs6f>acoburn: Oh, that's not an argument. That's just a problem. You solve that problem one funeral at a time.
<acoburn>ajs6f: yes, I know
* tolloid leaves10:24
<acoburn>ajs6f: in any case, their point is that not everything fits into Hadoop's Map-Reduce construct, but GAS is a very clever variation on that model
<ajs6f>acoburn: The API-X discussions often seem to me to edge very close to desiring to make inference part of the req-resp cycle. That demands totally new approaches.10:25
<acoburn>ajs6f: ahh, I see where you're going with this…
<ajs6f>acoburn: Not everything fits into any specialized computational model. The point of such models is that for the things that do fit into them, they work very well. What a dumb point— that's like saying that Torx screwdrivers suck because they do a crappy job as woodcarving chisels. Well, yes. _You're using them wrong._10:26
<acoburn>ajs6f: yes, exactly10:28
<ajs6f>acoburn: About four or five years ago I stopped having those conversations, not just at the office but in the rest of my life. I'm not an important man by any stretch of the imagination, but I'm not so unimportant that I have time to waste on that.10:29
* tolloid joins
<ajs6f>acoburn: Are you using Blazegraph? I know that some of the CLAW stuff uses it by default.10:30
<acoburn>ajs6f: my approach is to simply make a variety of tools available. If people want to use them, that's fine; if not, that's their choice
ajs6f: no, I'm using fuseki
<ajs6f>acoburn: Hm, I'll ask Diego about it. I'm pretty sure he uses it.10:31
<acoburn>ajs6f: I am not _committed_ in any strong way to fuseki — it's just that that's what I've got running, it works, and I don't have a compelling reason to change10:32
<ajs6f>acoburn: Yeah, I get that. I believe mikeAtUVa is using it too. I'm just looking for someone who might have tried the GAS API to get any feedback they have. I don't particularly want to get into the whole Spark/Elephas thing right now.10:33
<ruebot>ajs6f: fwiw, we're running it by default because it supposedly handles the most triples. dhlamb and the rest of were burned by mulgara, so we picked the "best" one at the time.10:48
ajs6f: there is this ticket too https://github.com/Islandora-CLAW/CLAW/issues/30
<ajs6f>ruebot: Mulgara is not even in the running, at all. That project hasn't been live since 2012.10:49
ruebot: Fuseki and Stardog are the choices I would put in a group with Blazgraph.
<ruebot>ajs6f++10:51
<ajs6f>AllegroGraph may also be in that group, but I haven't worked with it.
Ditto Sesame.
<ruebot>ajs6f: sesame is there, and allegro got crossed out because it isn't open source.10:52
<ajs6f>ruebot: I like diegopino's demand for "100% SPARQL compliance". Don't hold your breath, Diego. http://sparqlscore.com is the best benchmark I know. Even Jena isn't 100%, and Andy Seaborne is the freakin' editor of the SPARQL Query spec!10:53
<ruebot>ajs6f++10:54
Performance/Scale meeting here in a couple minutes: https://wiki.duraspace.org/display/FF/2016-07-18+Performance+-+Scale+Meeting10:58
* tolloid leaves11:02
<ruebot>https://wiki.duraspace.org/display/FF/Virginia+Tech+Results+-+Test+211:03
https://wiki.duraspace.org/display/FF/Virginia+Tech+Results+-+Test+3
https://wiki.duraspace.org/display/FF/Virginia+Tech+Results+-+Test+5
* tolloid joins11:05
* thomz leaves11:06
* peichman leaves11:16
* peichman joins11:17
<ajs6f>Just a random gentle reminder that code freeze for 4.6.0 is close-of-business _this_ Thursday, and the only remaining blocker is FCREPO-2065. We have a PR for that (#1059, thanks mesemus!) which will be merged soon. yay!11:42
* escowles joins11:45
@ruebot: i'm done with the notes for the perf. call: https://wiki.duraspace.org/display/FF/2016-07-18+Performance+-+Scale+meeting
and finished my testing of various versions of f4 with leveldb/postgres/mysql: https://wiki.duraspace.org/display/FF/Real+World+Performance#RealWorldPerformance-many_members
<ajs6f>Were you guys testing the 4.7.0 MODE5 branch? Or just the MODE4 mainline?11:46
ruebot: Can you think of a name put in the Islandora row for this: https://wiki.duraspace.org/display/FF/Release+Testing+-+4.6.012:03
?
<ruebot>ajs6f: me and/or whikloj12:05
escowles++
<ajs6f>ruebot:I'll put both of you and whoever is faster can delete his name.
<ruebot>ajs6f++
* whikloj is pretty fast at avoiding work
<ajs6f>GO!12:06
afk bbl12:10
* ajs6f leaves
<ruebot>http://islandora.ca/content/islandora-getting-technical-lead12:24
might be of interest for some folks here
* tolloid leaves12:32
* cmills joins12:33
* chadmills leaves12:36
* dchandekstark joins12:48
* anarchivist leaves13:06
* anarchivist joins13:13
* dchandekstark leaves13:14
* dchandekstark joins13:15
* tolloid joins13:24
* acoburn1 joins13:29
* acoburn leaves13:33
* dchandekstark leaves13:59
* dchandekstark joins14:01
* tolloid_ joins14:02
* dchandekstark leaves
* dchandek_ joins
* tolloid leaves14:03
* tolloid leaves14:04
* dchandek_ leaves14:05
* tolloid joins14:11
* arebenji leaves14:15
* jjtuttle leaves14:30
* jjtuttle joins14:31
* dchandekstark joins14:36
* dchandekstark leaves14:41
* tolloid leaves14:53
* tolloid joins14:54
* tolloid leaves14:55
* peichman leaves14:57
* peichman joins14:59
* ajs6f joins15:06
ruebot: when will you able to release a name?15:10
<ruebot>ajs6f: yeah, soon. waiting for paperwork to be signed.15:16
ajs6f: i assume you're talking about the tech lead :-)
ajs6f: ...and, yeah. i'm using MODE5 branch on the last two tests i've run
ajs6f: pretty sure escowles is doing the same.15:17
<ajs6f>ruebot: Yeah , that was my question. Okay, soon is good.
<escowles>ruebot: ajs6f: yep, testing mode4 and mode5, plus leveldb/postgres differences
<ajs6f>escowles:ruebot: Seeing any pattern of difference between MODE4 and 5?15:18
acoburn1: So you will be doing a Camel and Camel toolbox release for 4.6.0?15:19
<ruebot>ajs6f: MODE5 performs better... even with LevelDB
<escowles>ajs6f: generally seeing mode5 scale to larger numbers of objects
* peichman leaves15:20
<ajs6f>ruebot:escowles: Cool!
* peichman joins15:22
<acoburn1>ajs6f: yes, I'll have a new camel and camel-toolbox release for 4.6.015:23
<ajs6f>acoburn1: Okay, cool, I'll list that in the release Google doc.
acoburn1: Can I put you down to release-test that stuff?
<acoburn1>ajs6f: absolutely15:24
<ajs6f>acoburn1++
<ruebot>ajs6f: once we get in a better place with CLAW, I plan on doing the performance tests with the camel/async machinery hooked up to see how fcrepo, drupal, and camel perform/scale together
<ajs6f>ruebot: Cool, but not for this release, right?15:25
<ruebot>ajs6f: nope!
<acoburn1>ajs6f: I'm probably going to get a fcrepo-camel-toolbox release out sooner, too (peichman and mohammed contributed some code that they'd like to see in a release sooner)
<ajs6f>ruebot: Oh, wait— you're talking abot perf testing.
<ruebot>ajs6f: yeah :-)
<acoburn1>ajs6f: probably this week, then everything will be clear for a release cycle based on 4.6.0
<ajs6f>acoburn1: Okay, that's totally independent of release, right?15:26
<acoburn1>ajs6f: yes, it's entirely independent
ajs6f: I pushed really hard for that a while back, and I'm glad the projects have this sort of independency
<peichman>acoburn1++
<ajs6f>acoburn1: Okay, cool. Now if I could just figure out how to color cellsin Google Doc spreadsheets.
<acoburn1>s/independency/independence/15:27
<ajs6f>independencation
<acoburn1>ruebot: not "interdependence", O.N.A.N-style
<ruebot>acoburn1+++++++++++++15:30
<ajs6f>acoburn1: The audit triple store stuff— is that not going to work because of the same format change?15:36
<acoburn1>ajs6f: it's not going to work b/c we're not publishing properties any longer in the messages
ajs6f: it's something that will take ten minutes to fix15:37
<ajs6f>acoburn1: Can I leave it to you to file a ticket and so do in the next week or so?
<acoburn1>ajs6f: yes, I'll take care of it15:38
<ajs6f>acoburn1+++
acoburn1: That's the artifact fcrepo-audit, right?
<acoburn1>ajs6f: yes: org.fcrepo.camel/fcrepo-audit
<ajs6f>acoburn1: Okay, cool, I have a row in the table for that.15:39
<acoburn1>ajs6f++
<ajs6f>A successful Fedora release feels like : https://i.chzbgr.com/completestore/16/7/7/hxe1bqrBuUS7QIxQOlQBtQ2.mp415:45
<ruebot>ajs6f++15:52
* dchandekstark joins15:54
* dchandekstark leaves15:58
* escowles leaves16:16
* dwilcox leaves16:18
* github-ff joins16:56
[fcrepo-java-client] bbpennel opened pull request #14: Update readme with usage examples (master...update-readme) https://git.io/vKVyj
* github-ff leaves
<ajs6f>ruebot: have you got a link to the planning meeting on the 22nd for the im/ex work?16:57
* dwilcox joins16:59
* whikloj leaves17:00
* dwilcox leaves17:05
* dwilcox joins17:06
* acoburn1 leaves
* ajs6f leaves
* mikeAtUVa leaves
* dwilcox leaves17:08
<ruebot>ajs6f: yep; https://wiki.duraspace.org/display/FF/Design+-+Import+-+Export -- design page17:14
ajs6f: https://wiki.duraspace.org/display/FF/2016-07-22+-+Import+-+Export+Planning+Meeting
* peichman leaves17:16
* tolloid joins17:26
* jrgriffiniii joins17:44
* tolloid leaves18:11
* jrgriffiniii leaves20:12
* jrgriffiniii joins20:48
* dwilcox joins20:49
* dwilcox leaves20:50
* github-ff joins21:27
[fcrepo-java-client] acoburn closed pull request #14: Update readme with usage examples (master...update-readme) https://git.io/vKVyj
* github-ff leaves
* travis-ci joins21:31
fcrepo4-exts/fcrepo-java-client#42 (master - 093653a : Aaron Coburn): The build passed.
Change view : https://github.com/fcrepo4-exts/fcrepo-java-client/compare/4abdd78fc031...093653a30679
Build details : https://travis-ci.org/fcrepo4-exts/fcrepo-java-client/builds/145707815
* travis-ci leaves
* mikeAtUVa joins23:27
* mikeAtUVa leaves23:40

Generated by Sualtam