Perl 6 - the future is here, just unevenly distributed

IRC log for #webwork, 2012-10-15

| Channels | #webwork index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:55 mgage rbeezer:  paul seeburger reported that hosted2 was down Thursday night -- or at least running really really slowly.  He said it improved by about 11:30 and I rebooted the machine when I got to Fitchburg -- around midnight Thursday night.   Have you had difficulties since?
00:55 rbeezer mgage: thanks for the info.  No problems since, we are now on Fall Break ;-)
00:56 rbeezer Paul's report seems consistent with what my students said when I polled them in class
00:56 mgage k -- we had a good code camp.   David and Peter got a lot of work done on instructor interfaces.  we had some good discussions with people coming in from WHFreeman -- including Sam Hathaway
00:57 rbeezer glad to hear code camp went well,
00:57 mgage I don't know what went wrong on hosted2 but it has the symptoms of a stuck child -- that was probably cured by the regular reboot at midnight
00:57 mgage it was short -- it went well.
00:57 rbeezer I've written all my problems for the semester (except the final section, which is new),
00:58 mgage wow -- you're way ahead.
00:58 rbeezer looks like next fall, I can try to clean them up to make them usable publicly
00:58 rbeezer Busy working on the textbook, so I blasted through the WW problems
00:59 rbeezer Guess hosted2 just needed a kick
00:59 mgage very nice.  I got a little further on improving the sequential/compound problem support -- I sent some files to JT but I can send them to you also if you like -- they can still use more polishing but they are usable.
00:59 mgage yep
00:59 mgage these computers need that every now and then
01:00 rbeezer It'd be a good time for me to have a look, I have several more days without any classes
01:00 mgage ok -- sent
01:01 rbeezer thanks, I'll put them on the stack  ;-)
01:01 mgage the syntax is getting cleaner -- but it can be cleaned up even more
01:01 rbeezer xml  ;-)
01:01 rbeezer (one-trick pony)
01:01 mgage JSON :-)
01:01 rbeezer OK!
01:02 mgage ttyl
01:02 rbeezer k
02:16 Paul_Pearson joined #webwork
13:17 aubreyja joined #webwork
13:18 mgage hi
13:23 jpgravel joined #webwork
13:23 mgage hi
13:28 jpgravel hi mr gage
13:28 jpgravel did you get my private message last friday?
13:28 mgage hi -- we're just back from the code camp at Fitchburg State  -- I'm catching up on my email
13:29 Paul_Pearson joined #webwork
13:29 mgage hang on I'm just reading it
13:29 jpgravel nice did a nice trip?
13:29 jpgravel ok
13:30 mgage yes -- we got quite a lot of work done.  I'm delighted that you and your students are interested in helping out with WeBWorK development
13:31 jpgravel good, so I want have some direction to get involved and be efficient with the on going development
13:31 mgage the short answers to your questions on the ww-devel list:
13:31 jpgravel but you may be interested in some background from our project first?
13:32 jpgravel ww-devel? do you mean the mail list arcive?
13:32 mgage no-one working on a Maple T.A. to WeBWorK question translator to my knowledge.  Gavin LaRose (glarose@umich.edu) has used both Maple T.A. and WeBWorK for placement tests but I don't think he wrote a translator
13:32 mgage yes
13:34 mgage There have been a number of requests for a friendly question editor and in fact a couple of people worked on that project during this code camp -- I'll reply to your email and give a few more details -- I don't think there is a definitive "friendly" editor yet -- so there is plenty of room for development
13:34 jpgravel good, in fact a translator would be too hard to get done in a reasonnable time (and budget). mta2ww shall be closer to a translation helper than a translator
13:35 jpgravel if you see what i mean
13:35 jpgravel but concerning our plans for questions and equations, what is the current advancement?
13:35 mgage As djun mentioned to you we have plugged in dragmath as an option for equation editing.  I would like to see some more options and there are others who are also looking at javascript or HTML5 solutions.  again no definitive solution yet
13:37 mgage ok -- those are the short answers -- I'll try to write something more comprehensive this evening and I'll put it on the development mail list (and copy you).  mostly it will be a list of people who have indicated interest in working on question and equation editors
13:37 mgage what can you tell me about your project?
13:38 jpgravel The project was initiated by Louise Pellerin from the Céget de Chicoutimi (a college in quebec) to use webwbork instead of Maple TA for homework assignement
13:38 jpgravel she packed that idea into a project and asked the French College association in Canada to get some budget
13:38 jpgravel and she got it
13:39 jpgravel so she hired me as a project management and technical consultant to coach her students
13:39 mgage great.
13:39 jpgravel the project have three goals:
13:39 jpgravel Ceget I meant Cégep
13:40 mgage I suggest that the two of you write a brief note to Gavin LaRose just to get the benefit of his insights after making a similar switch.
13:40 jpgravel 1) Translate the WebWork backend to french
13:40 jpgravel ok, nice suggestion, who is Mr Larose?
13:41 jpgravel 2) Make a Maple TA to WebWork translatio helper
13:41 jpgravel 3) Develop a user firendly question editor (including an equation editor)
13:42 mgage Gavin is a mathematician at U. of Michigan,  he has been working with WW for many years and for project NExT for even longer:  http://www.math.lsa.umich.edu/people/facultyDetail.php?uniqname=glarose
13:44 jpgravel thank you for your insights, i'll have a chat with Mr Larose and do some reading in the ww-devel section of the web site
13:44 jpgravel have a nice day, Im subscribed to the ww-devel list so I'm waiting for your mail.
13:45 mgage So we have made significant progress in localizing the basic code of WeBWorK so that it can be translated into many languages but there is still a substantial amount of work left to do.  (see  some of the blog posts at http://webwork.maa.org/planet/    in particular:   http://michaelgage.blogspot.com/  and sebastian's posts at http://thales.math.uqam.ca/~labbes/blogue/categorie/webwork/
13:47 jpgravel ok we got some insight of the localisation system being developped and tried it
13:47 jpgravel As I said to djun, Louise Pellerin is in contact with Mr Labbé
13:48 jpgravel We will see what we can do on that
13:49 jpgravel I got a load of reading to do now ;)
13:50 mgage the actual questions are probably best just rewritten -- I don't see an easy way to automatically translate them.  Stephanie Lanthier was also involved.  I can also find names of other participants from Cegep colleges both english and french.  There has been quite a lot of interest in WeBWorK in the Montreal area and north -- I grew up in Canton, NY on the St. Lawrence so I'm happy about that.
13:50 mgage sounds good -- I need to take off as well.  Talk to you later (ttyl)
13:50 jpgravel ok nice, goodbye
13:50 jpgravel and thanks again
13:51 mgage np
14:25 Paul_Pearson joined #webwork
15:40 mgage_ joined #webwork
15:41 mgage_ hi
15:41 mgage_ aubreyja: hi
15:55 Paul_Pearson joined #webwork
16:29 mgage_ joined #webwork
16:59 mgage_ joined #webwork
17:00 aubreyja Hi - back - how was the code camp?
17:01 mgage_ it went pretty well I think -- particularly for such a short time.  We had some IT difficulties the first day getting the wireless and projectors to work but things settled down
17:01 mgage_ Peter and David got quite a bit done on infrastructure for instructor pages
17:02 mgage_ I talked to the Laura Judge and Terri Ward at Freeman for some time as well as to Sam Hathaway
17:03 mgage_ Sam said we can start asking him to  participate again -- he is particularly interested in getting a stand alone pg installation running to serve as a back end for some of their LMS
17:06 mgage_ we also talked to WPI students -- who worked a little on simple input.  Neil Hefferenan from WPI also came up and gave us a quick introduction to his creation ASSISTments (which got mentioned in the NYT magazine last week)   Bruce Romano (he is at FSU and was at one of the consultant workshops ) also came and worked on simple input
17:06 mgage_ Christine Kayastha -- she was at the June code camp -- also came and helped out with Peter and David's work.
17:07 mgage_ I worked on Davide's compound problem structure a bit and sent the results off to John Travis.  Davide and I talked about possible future enhancements.  He also has a prototype simple input format.
17:10 aubreyja joined #webwork
17:10 aubreyja Sorry - my computer connected to the vpn and my irc client disconnected.
17:11 aubreyja Glad to hear a lot of progress was made.
17:11 mgage_ pretty good for two days total.
17:11 aubreyja It will be good to have Sam participating again too
17:11 mgage_ yes
17:11 aubreyja Did the chagnes get pushed to webwork-dev?
17:11 mgage_ I think we made some useful connections.  I think if we can get the 501© running that WHFreeman will be even more interested
17:12 aubreyja Yes, I sent out a meeting invitation to try to organize something with Davide and his colleague at mathjax, but I think we were all too busy last week
17:12 mgage_ the changes to WW code got pushed to mgage/webwork2-dev from Peter and David's git hubs.  I haven't pushed them to the openwebwork site yet -- this evening or later this afternoon.
17:13 aubreyja Did you talk with Davide at all about his interest in possibly having a joint 501(3)©?
17:13 mgage_ we need to keep at it -- I've been pretty busy with projects here at the U. of R so I haven't been much help on that front.  does it still look doable?
17:14 mgage_ Yes -- he is interested -- he thinks it would be a good idea for MathJax
17:14 aubreyja Yes, they are definitely interested - I'll just do a new meeting invitation this afternoon
17:14 mgage_ ok -- I'll try to make this -- Tuesdays and Thursdays are pretty packed for me during the day until 5pm
17:15 aubreyja I can generally do MWF afternoons too - we'll find something. weekend if we need too
17:15 mgage_ mon-wed I have a class 3:25--4:40 but the rest of the day is pretty free
17:17 aubreyja btw, our performance issues came back somewhat again, and it seems that gateway quizzes are the trigger (and midnight deadlines, which I've gotten them to change)
17:17 aubreyja but, one idea I got interested in is making the gateway code more efficient
17:17 mgage_ I've talked to Jeff Bigham, the HCI guy here, about giving us a general talk about accessibility issues and guidelines/tools for addressing it.  I was thinking of setting up a google+ hangout.  Can you help me make a list of people who we should invite/ who would be interested?  I could also send out a query on the newsletter or forums if you think that is a good idea
17:18 aubreyja Ah yeah, that would be great
17:18 Paul_Pearson joined #webwork
17:18 aubreyja Vicki and Flora for sure
17:19 mgage_ for performance testing -- our attempts to use Dprof long ago were pretty frustrating but I understand NYTDprof (a new version of dprof developed by the new york times is considerably better)
17:19 aubreyja We've probably had people ask about this on the forums - we should invite them too.  I'd also invite David G, Peter, etc. - anybody working on the new UI stuff
17:20 aubreyja If I recall correctly, the gateway code creates an array of separate pg objects, one entry for each problem in the quiz
17:20 mgage_ I think that is right
17:20 aubreyja so a lot of things get compiled into separate safe compartments over and over
17:22 aubreyja it seems like it should be possible to create one uber environment to share math object code for example
17:22 mgage_ yes -- we haven't been caching the safe compartment as much as we did originally -- processors got faster and it was complicated maintaining everything.  On the other hand more of the code is now in .pm modules where it is cached by mod_perl
17:23 aubreyja yeah, I don't understand how that works right now. Adding Value.pl to the cache list helps a lot, but it also makes the size of the apache processes blow up
17:24 mgage_ it's possible to do this, then transfer the namespace to the current instance of Safe --- we were doing that a lot although it's complicated.  The mechanisms are still there in the Translator.pm file.  Some of the use of those subroutines has been commented out in PG/Local.pm
17:24 aubreyja Davide is right that a lot of PG code gets cached now, it even looks like some math object stuff, but adding Value.pl has a big effect, so something is going on I don't' understand
17:25 mgage_ not as much code is being cached as Davide remembers, including probably Value.pl -- the .pm files are cached.
17:26 mgage_ I would like to see some actual profiling before we decide which are the performance bottle necks.  guessing at those by dead reckoning is error prone
17:26 aubreyja But on the other hand, I'm seeing the Value:: pm files getting reloaded over and over.
17:26 aubreyja Yes,
17:27 aubreyja I've got a dev server I'm going to set up (webwork-dev.missouri.edu), and I'm going to add Devel::NYTProf, etc., and let it run with that for a while
17:27 mgage_ and I'm puzzled about why that would occur.  -- at least for real -- they might read the use Value line but then notice that it is already loaded.
17:27 mgage_ that sounds good.
17:28 aubreyja Well, our performance errors were often of the form 'Can't reload Parser::Legacy'
17:29 djun joined #webwork
17:30 mgage_ you might have had more children spawned than can fit in memory -- that usually leads to a performance hit.  I found keeping the number of children low actually speeded up performance in most cases. -- it dramatically reduced swapping
17:55 aubreyja Sorry - student - back.  I'll check on this.  We had to up MaxClients from 150 to 256 and then again to a higher number that I don't know off hand because we actually had more than that many concurrent users.  Caching Value.pl made that error go away, but then under really high load we were swapping because the apache processes were a lot bigger.
17:57 mgage_ my experience was that keeping maxclients low helped -- a student might wait a second for a response to a request but that was better than the result when there were too many children and a lot of swapping taking place.
17:57 mgage_ but real profiling data would eliminate a lot of the guess work
17:59 aubreyja yeah, it's a balance - the first night we exceeded MaxClients students and instructors were emailing saying they couldn't connect for a long time, but when I logged into the server, it was humming along, no swapping, plenty of available ram, etc.  Everything looked fine and it took me a while to realize what that meant.  Now we're probably erring on the other side of that equatoin.
18:15 mgage_ usually if max clients is too low the worst that happens is a few requests are dropped.  other requests are queued for a brief time before being answered -- I think these are hard to distinguish from network delays
18:15 mgage_ aubreyja: hi
18:17 mgage_ k -- I'm off to classes -- back arougn 5pm
19:09 Guest60884 joined #webwork
20:37 djun joined #webwork
20:41 Paul_Pearson joined #webwork
21:22 mgage_ joined #webwork
22:07 djun joined #webwork
22:28 mgage_ joined #webwork
22:54 djun joined #webwork
23:17 Paul_Pearson joined #webwork

| Channels | #webwork index | Today | | Search | Google Search | Plain-Text | summary