2010-10-29T01:09:37 *** tildeequals has joined #evergreen 2010-10-29T01:54:16 *** rsinger has quit IRC 2010-10-29T02:17:52 *** rsinger has joined #evergreen 2010-10-29T02:46:43 *** tildeequals has quit IRC 2010-10-29T02:54:43 *** rsinger has quit IRC 2010-10-29T03:17:27 *** rsinger has joined #evergreen 2010-10-29T03:23:03 *** tildeequals has joined #evergreen 2010-10-29T03:34:09 *** rsinger has quit IRC 2010-10-29T03:35:31 *** rsinger has joined #evergreen 2010-10-29T04:02:15 *** rsinger has quit IRC 2010-10-29T04:25:01 *** rsinger has joined #evergreen 2010-10-29T04:57:25 *** natschil has joined #evergreen 2010-10-29T05:01:26 *** natschil has quit IRC 2010-10-29T06:04:43 *** eby has quit IRC 2010-10-29T07:05:03 *** shopkins has joined #evergreen 2010-10-29T08:08:30 *** granitize1 has quit IRC 2010-10-29T08:08:51 *** granitize has joined #evergreen 2010-10-29T08:21:07 *** ebyr has joined #evergreen 2010-10-29T09:07:36 *** Callender has quit IRC 2010-10-29T09:08:01 *** Meliss has joined #evergreen 2010-10-29T09:10:04 *** ebyr has quit IRC 2010-10-29T09:10:22 *** eby has joined #evergreen 2010-10-29T09:10:25 *** lisppaste has joined #evergreen 2010-10-29T09:25:52 *** jenny1 has joined #evergreen 2010-10-29T09:27:31 *** bshum has joined #evergreen 2010-10-29T09:27:32 *** sfortin has joined #evergreen 2010-10-29T09:31:39 sure! 2010-10-29T09:32:19 where everyone cares 2010-10-29T09:32:20 eby: so Perl and Python have the CStoreEditor which gives you a nice convenicience wrapper rather than calling open-ils.cstore directly 2010-10-29T09:32:38 yeah i think once you have a good opensrf client 2010-10-29T09:32:54 getting a bunch of wrappers should be easy for anyone to add on 2010-10-29T09:32:56 I was suggesting that a PHP EG connector might want that, but yes - a good OpenSRF client is a prereq :) 2010-10-29T09:33:05 cause php users LOVE CONVENIENCE 2010-10-29T09:33:28 We'll have to make sure we switch the order of arguments in the methods randomly 2010-10-29T09:33:50 okay, but cstore isn't exposed via the public gateway/translator, right? 2010-10-29T09:34:00 jeff: that is correct 2010-10-29T09:34:11 so this would be for php that's running XMPP against the private router? 2010-10-29T09:34:24 jeff: that is correct 2010-10-29T09:34:41 okay, that's why i wasn't following over in #code4lib. i've never been tempted to do that. :-) 2010-10-29T09:34:47 you're probably sane 2010-10-29T09:35:13 are editing/updating of data done over the public or only private 2010-10-29T09:36:06 eby: the "public" router exposes services whose methods enforce permission checks, etc. 2010-10-29T09:36:24 * eby has it on his list to get opensrf running again next week 2010-10-29T09:36:50 multibox opensrf. saw the docs were kind of bare with the multiserver setup 2010-10-29T09:36:58 eby: the "private" router has things like cstore, which does not do those kinds of checks, and is not exposed on the "public" router, but is used by many of the services that ARE on the public router. 2010-10-29T09:37:30 so public router services do various checks and make calls to the private router services 2010-10-29T09:37:36 eby: the staff client uses the services on the public router, so yes... data can be edited/updated via the "public" router, given appropriate permissions 2010-10-29T09:38:02 eby: right. the private router has all the services, the public router has the "safe" services. 2010-10-29T09:38:22 and someone else can correct me if i've over-simplified or gotten something completely wrong. :) 2010-10-29T09:41:08 jeff: correctamundo. and the public services talk to the private services privately, so the only data that flows from the server to clients (like the staff client or web browser) is supposed to be public data 2010-10-29T09:41:27 some auditing of that needs to be done, mind. 2010-10-29T09:42:23 maybe opensrf-over-http should be opensrf-over-https... 2010-10-29T09:42:37 i'm not sure I follow the "public data" concept there. The staff client fetches plenty of what I'd consider "private" data via the "public" router. I don't think the "router" public/private and "data" public/private comparisons are equivalent. 2010-10-29T09:43:12 jeff: no, they're not equivalent 2010-10-29T09:44:05 I'm muddying the waters by not focusing, should just bear down on next job 2010-10-29T09:45:07 i understood that as not exposing the private methods 2010-10-29T09:45:08 public data vs. private data is something that probably requires a further distinction 2010-10-29T09:47:45 *** Callender has joined #evergreen 2010-10-29T09:59:37 public API versus internal API 2010-10-29T10:01:48 dbs: any objections to me updating the gentoo prereq page? i see no one was updating it so you deleted 2010-10-29T10:02:45 eby: no, if you have a working implementation on gentoo then feel free to bring it back to life (and Makefile.install as well, for that matter) 2010-10-29T10:06:20 * dbs really needs to write up Fedora 13 install instructions, just in time for Fedora 14 2010-10-29T10:13:07 eby: we're running a simple multibox setup; ejabberd + one set of opensrf services on one box, another set of opensrf services on another box, and memcached on a third box 2010-10-29T10:14:13 full opensrf install on both services boxes 2010-10-29T10:14:23 as long as you only have the opensrf router running on one box, then I think it's pretty straightforward. 2010-10-29T10:14:37 i believe this is what is called a "brick" in the standard nomenclature (each brick gets its own router - but I could be entirely wrong about that) 2010-10-29T10:15:00 we haven't had to deal with that level of complexity 2010-10-29T10:15:39 eby: yep, full opensrf install on both, although we mount the /openils directory via a read-only NFS share on the second box so that the config can be shared 2010-10-29T10:15:59 this means that your logs & pids and such have to go somewhere other than /openils 2010-10-29T10:22:38 *** jenny1 has quit IRC 2010-10-29T10:30:02 * berick confirms dbs' "each brick gets its own router" 2010-10-29T10:35:02 *** jenny1 has joined #evergreen 2010-10-29T10:39:26 berick: what was the situation where one might want to pierce the brick wall, so to speak? You know what I'm talking about? 2010-10-29T10:44:24 phasefx: yes. state-full requests (via osrf translator), where subsequent calls are routed to a different apache server 2010-10-29T10:44:37 the routed calls have to find their way back to the original brick 2010-10-29T10:45:10 since the client is communicating w/ a specific drone 2010-10-29T10:45:38 so, for example, doing something within a transaction? 2010-10-29T10:46:24 like with open-ils.pcrud.transaction.begin and .commit 2010-10-29T10:47:48 phasefx: exactly 2010-10-29T10:48:11 and the solution there is to have OSRFTranslatorCacheServer on each apache pointing to the same server? 2010-10-29T10:48:24 (and thus breaking the brick concept) 2010-10-29T10:48:46 or 2010-10-29T10:48:49 the solution is make sure jabber on every brick support server-2-server communication and is able to location remote jabber server 2010-10-29T10:48:53 have server-server jabber.. yeah 2010-10-29T10:49:14 * phasefx groks again, gracias 2010-10-29T10:49:21 you need the same cache server as well, though 2010-10-29T11:00:46 *** r123 has joined #evergreen 2010-10-29T11:03:10 *** granitize has left #evergreen 2010-10-29T11:09:19 *** granitize has joined #evergreen 2010-10-29T11:16:04 *** granitize has left #evergreen 2010-10-29T11:23:37 *** granitize has joined #evergreen 2010-10-29T11:43:34 *** mmorgan has joined #evergreen 2010-10-29T11:47:43 *** rickd_ has joined #evergreen 2010-10-29T11:53:47 *** jenny1 has quit IRC 2010-10-29T12:09:50 *** brendan2 has joined #evergreen 2010-10-29T12:11:43 *** jenny1 has joined #evergreen 2010-10-29T12:36:57 My brain refuses to stop trying to over-examine possible issues with dynamic weighting on circ matrix stuff. That and it keeps telling me to do the same to holds while I am at it. <_< 2010-10-29T12:39:07 *** brendan2 has quit IRC 2010-10-29T12:39:27 *** jamesrf has joined #evergreen 2010-10-29T12:42:18 *** Meliss has quit IRC 2010-10-29T12:44:18 *** Meliss has joined #evergreen 2010-10-29T12:47:09 *** brendan2 has joined #evergreen 2010-10-29T12:49:20 *** jenny1 has quit IRC 2010-10-29T12:51:05 *** jenny1 has joined #evergreen 2010-10-29T13:07:59 tsbere: That's okay. Someday we'll offer you an award certificate for your efforts. It might be printed on a napkin, but consider it heartfelt thanks ;) 2010-10-29T13:08:29 Good news is I think I have a workable, easy to implement sorting algorithm. <_< 2010-10-29T13:08:56 anyone converted their indexes from GIST to GiN? 2010-10-29T13:10:39 jamesrf: played with adding a GIN index and dropping a GIST index a while back, but not in production 2010-10-29T13:12:16 i've put one on on the keyword and title indexes on a non production box and it is much faster 2010-10-29T13:12:37 explain analyze select count(*) FROM metabib.keyword_field_entry where index_vector @@ to_tsquery('doors'); takes 58546.728 ms on prod 2010-10-29T13:12:54 same query on non-prod database with GIN is 372 ms 2010-10-29T13:14:29 That would qualify as much faster :) 2010-10-29T13:14:32 4468278 records 2010-10-29T13:15:22 we haven't had any performance bottlenecks, thus no GIN in production. I recall when first looking at them that there were some drawbacks, but that was pre 8.3 - still pretty raw 2010-10-29T13:18:19 bugs like http://archives.postgresql.org/pgsql-bugs/2009-11/msg00100.php worried me a bit 2010-10-29T13:20:04 even in 8.4.5 "Cases involving AND/OR combination of several GIN index conditions didn't always give the right answer, and were sometimes much slower than necessary." 2010-10-29T13:21:16 but if you're running 8.4.latest or 9.0, then it seems worth trying out to see if it actually affects the OPAC search response time 2010-10-29T13:32:03 *** jenny1 has left #evergreen 2010-10-29T13:42:15 hmm those are a bit icky 2010-10-29T14:03:03 is there a preference of apache worker or prefork with eg? 2010-10-29T14:05:01 I think prefork is required 2010-10-29T14:05:14 threads begone 2010-10-29T14:06:15 I haven't tried the "use a mpm that allows for per-vhost user/group settings" trick to make apache play nicer with other sites, but I suspect that it won't be as happy. 2010-10-29T14:16:02 Is it me, or is it literally impossible to guarantee that you have all of your circ matrix entries for "don't circulate" in the hold matrix so that items that won't circ also won't be available for holds? 2010-10-29T14:22:15 *** shopkins has quit IRC 2010-10-29T14:24:32 tsbere: could you just set a rule for all patrons to not circulate anything and then override it with the circmods they can circulate? 2010-10-29T14:25:53 er no sorry i think i misunderstood your problem 2010-10-29T14:32:00 jamesrf: I am saying "I can create a circ rule that says an item won't circulate such that I can NOT create a hold rule that says it isn't holdable, leading to a situation wherein I can place a hold on an item that isn't allowed to circulate" 2010-10-29T14:32:49 ah yes 2010-10-29T14:33:32 I see that as a problem 2010-10-29T14:37:34 yes i see, it is also a problem in legacy circ scripts 2010-10-29T14:37:51 although we have never actually encountered it in production as far as i know 2010-10-29T14:38:11 the legacy scripts do check for the circulate = false on the copy 2010-10-29T14:38:25 I am not looking to fix legacy scripts. I am looking to provide an option for fixing it in the database, even if it defaults to "don't check that". 2010-10-29T14:39:02 maybe add a trigger that updates the hold matrix when you update the circ matrix? 2010-10-29T14:39:12 Won't work 2010-10-29T14:39:33 The circ matrix can check things the hold matrix can't and vice-versa. There isn't any one to one relation. 2010-10-29T14:41:09 * tsbere almost wants to create a third way of doing circ/hold stuff and call it indb2 or something. <_< 2010-10-29T15:06:10 tsbere: we allow staff to place holds on behalf of patrons for items that do not circulate. 2010-10-29T15:06:34 (just as a data point :) 2010-10-29T15:07:18 jeff: That.....seems odd. Perhaps I should add column to the hold matrix saying "check circulate?" instead of a global flag? 2010-10-29T15:09:04 i'm not yet familiar enough with indb to comment there. 2010-10-29T15:09:37 our use case is ILL microfilm. we need to notify the patron that it's here, but they do not circulate it, they just use it in-building. 2010-10-29T15:14:15 i could see use using similar setup 2010-10-29T15:14:22 we def have circ but no hold 2010-10-29T15:17:06 Circ without hold is easy. Hold without circ is a tad odd IMO, especially when you can configure a "won't circ" that can't be configured as a "won't hold". 2010-10-29T15:17:28 Although jeff provided a decent example of a use case 2010-10-29T15:18:18 * dbs would argue that that microfilm should still circ, but with a four-hour interval or something like that 2010-10-29T15:20:41 * tsbere would agree that should be the case, but isn't about to argue that it has to if the library wants to do something differently 2010-10-29T15:21:26 * dbs is not going to argue that it has to either. ergo the use of "should" 2010-10-29T15:21:49 i don't know how we handle when the patron comes in -- since we don't check out the materials, we probably cancel the hold, which is a weird way to handle it. 2010-10-29T15:31:45 *** pmplett has joined #evergreen 2010-10-29T15:44:01 So far so good, my "associate weights with org units" code appears to work. Needs a foreign key relationship or two, but I will get there. 2010-10-29T16:00:31 *** Meliss has quit IRC 2010-10-29T16:05:03 * moodaepo thinks an option for "off line" mode for opac would be neat to have during upgrades the opac. 2010-10-29T16:05:35 moodaepo: I think that's called "VuFind" 2010-10-29T16:05:38 oops should have been "during upgrades." 2010-10-29T16:06:04 *** granitize has left #evergreen 2010-10-29T16:06:53 Hrm. Anyone here know of a way to prohibit certain items (say, by circ modifier) from transiting to fulfil a hold? 2010-10-29T16:07:11 dbs: True. We do run VuFind for our Aleph instances...might have to look into the EG connector status 2010-10-29T16:07:15 jeff: You're asking an in-db question or scripted? 2010-10-29T16:07:28 bshum: circ scripts. 2010-10-29T16:07:38 with all of their flexible legacy goodness 2010-10-29T16:07:41 jeff: Ah, then I can't help you :) 2010-10-29T16:08:13 moodaepo: Isn't Tasmania using VuFind with their EG? (can't recall if they were 1.4 or 1.6) 2010-10-29T16:08:50 And I think Pennsylvania is going to use VuFind with 2.0, but the folks there are working on the connectors last I heard. 2010-10-29T16:09:44 bshum: I think they use 1.4 2010-10-29T16:10:22 *** pmplett has quit IRC 2010-10-29T16:10:38 *** pmplett has joined #evergreen 2010-10-29T16:11:42 Is there a good way (in 8.4 postgres) to get a list of permission group ancestors with a "proximity" like the org unit proximities? 2010-10-29T16:12:37 *** pmplett has quit IRC 2010-10-29T16:12:53 *** pmplett has joined #evergreen 2010-10-29T16:18:35 *** bshum has quit IRC 2010-10-29T16:21:06 *** bshum has joined #evergreen 2010-10-29T16:21:28 tsbere: you want a permission.perm_grp_proximity table like actor.org_unit_proximity? 2010-10-29T16:22:10 phasefx: I don't care about group to group, just how far down the tree I am (I do not forsee ever caring about the proximity of a staff subgroup to a patron subgroup, for example) 2010-10-29T16:22:17 Or perhaps more "how far back UP the tree I am" 2010-10-29T16:23:20 I see a "grp_ancestors" function, but that just returns a list, nothing to say what the tree order of that list would be 2010-10-29T16:23:50 * phasefx could write a stored procedure to do it 2010-10-29T16:24:21 So could I, but for all I know there is a function that will just do it for me, effectively, in 8.4. :P 2010-10-29T16:24:30 yeah, I have no idea :) 2010-10-29T16:24:52 Although since I care about it for two possible places, if there isn't a way to just do it a new stored procedure is probably a good idea. 2010-10-29T16:25:57 * tsbere could also just add a "depth" field to the permission tree with appropriate triggers 2010-10-29T16:28:01 something like "on insert or update, if the new row has a parent, the depth of the row becomes one more than the depth of the parent, else the depth becomes 0" 2010-10-29T16:28:51 Although cascading that down to children would be problematic........ 2010-10-29T16:29:51 * tsbere would just ignore this problem if he didn't have to solve it to code things the way he wants to 2010-10-29T16:33:41 tsbere: WITH RECURSIVE? 2010-10-29T16:33:48 http://www.postgresql.org/docs/8.4/static/queries-with.html 2010-10-29T16:33:57 Hey, I am reading that page now! 2010-10-29T16:34:06 *** jamesrf has quit IRC 2010-10-29T16:41:01 I think I have something that will work 2010-10-29T16:42:54 Happy Friday! The wiki has been updated with Release Notes for 2.0 (although OPAC notes are not yet up as OPAC dev is in some flux). http://evergreen-ils.org/dokuwiki/doku.php?id=feature_list_2_0 2010-10-29T16:43:14 gdunbar++ 2010-10-29T16:43:38 slipscomb++ (she did the heavy lifting on these) 2010-10-29T16:46:44 tsbere pasted "grp_*_depth" at http://paste.lisp.org/display/116030 2010-10-29T16:46:51 Any thoughts on those two? 2010-10-29T16:56:58 *** samsung has joined #evergreen 2010-10-29T17:04:50 *** mmorgan has left #evergreen 2010-10-29T17:04:59 *** bshum1 has joined #evergreen 2010-10-29T17:08:11 *** bshum has quit IRC 2010-10-29T17:14:25 *** brian_f has joined #evergreen 2010-10-29T17:16:24 *** bshum1 is now known as bshum 2010-10-29T17:22:57 tsbere annotated #116030 "Function Form" at http://paste.lisp.org/display/116030#1 2010-10-29T17:24:52 dbs or phasefx or anyone else who cares: Any thoughts or issues you see with those two? Or desire to make them part of trunk before I make code that needs them? 2010-10-29T17:27:33 tsbere: I glanced, but my brain is already melting. eeevil is away today but I bet he'll stop in tonight or this weekend 2010-10-29T17:36:27 *** gdunbar has quit IRC 2010-10-29T17:36:45 use /nick cheesemonger 2010-10-29T17:40:00 *** samsung has left #evergreen 2010-10-29T17:41:17 *** cheesemonger has joined #evergreen 2010-10-29T17:42:43 *** sfortin has quit IRC 2010-10-29T17:57:02 *** bshum has quit IRC 2010-10-29T18:23:02 *** pmplett is now known as pmpafk 2010-10-29T18:45:47 *** cheesemonger has left #evergreen 2010-10-29T19:04:13 *** jamesrf has joined #evergreen 2010-10-29T20:16:33 *** jamesrf has quit IRC 2010-10-29T20:28:35 * phasefx did a search for "mines of moria" at target.com, and got back results for "miles of maria" (and read it as "malaria") 2010-10-29T21:02:01 *** jamesrf has joined #evergreen 2010-10-29T21:19:52 *** brian_f has quit IRC 2010-10-29T21:39:30 *** CarlSagan has joined #evergreen 2010-10-29T21:40:16 *** tildeequals has quit IRC 2010-10-29T21:42:17 good evening gentlemen, gentlewomen! i've got a question concerning openSRF. when i try to configure, i get the error message "for library containing mc_req_free... no 2010-10-29T21:42:24 configure: error: ***OpenSRF requires memcache development headers 2010-10-29T21:43:07 i know where to get memcached and i already have it, but where is openSRF looking for this file? does anybody know? 2010-10-29T22:07:09 CarlSagan: what linux distro are you trying this on? the Makefile.install process should install the required development libs. 2010-10-29T22:13:05 trying this on arch linux kernel 2.6.35. 2010-10-29T22:13:19 maybe that's the problem? 2010-10-29T22:15:19 jeff: i know all the documentation references ubuntu and debian packages... 2010-10-29T22:16:19 since there's no Makefile.install target for arch, what distro did you provide as an argument when installing the pre-reqs? 2010-10-29T22:17:37 i'm not very familiar with arch / pacman. you'll probably want to pick a distro, see what pre-reqs are installed for that distro, and determine (and manually install) equivalent packages for arch. 2010-10-29T22:18:47 okay. arch is similiar to debian in a lot of ways, so that will probably be the best way to go... 2010-10-29T22:19:23 if you've installed a memcached package, you probably also want a memcache-dev or similar package. 2010-10-29T22:19:31 thank you for your advice. 2010-10-29T22:19:43 okay! i'll install that too. 2010-10-29T22:24:14 CarlSagan: libmemcache and libmemcache-dev, specifically 2010-10-29T22:24:40 note that newer releases of OpenSRF use a different client library, libmemcached 2010-10-29T22:25:00 okay. it looks like i can get that one from pacman. 2010-10-29T22:25:37 libmemcached, that is. 2010-10-29T22:26:22 gmcharlt: thanks. i was trying to remember/find the demarcation between libmemcached and previous. 2010-10-29T22:30:33 * tsbere can't figure out what this strict_ou_match thing is supposed to do in the hold matrix 2010-10-29T22:34:17 Oh, wait, the weight is better when lower. Ok, that makes a tad more sense...... 2010-10-29T22:57:23 okay! this is starting to come along now! thank you guys for your help. 2010-10-29T22:57:38 * tsbere understands it more now, but thinks something is wrong with the picture on the strict_ou_match stuff..... 2010-10-29T23:06:47 *** pmpafk has quit IRC 2010-10-29T23:20:16 * tsbere has now declared the entire indb hold matrix check set horribly broken and unusable in the current state 2010-10-29T23:30:15 *** pmpafk has joined #evergreen