summary: today discussion was about remaining OM issues (not much) and initial OM performance results - we did not have enough time to discuss WSDL processing.
please make sure to visit and add topics to agenda for next week chat at: http://wiki.apache.org/ws/ChatAgenda
[10/13/2004 8:11 AM] <alek> shoud we get started? [10/13/2004 8:12 AM] <Srinath> hi all [10/13/2004 8:12 AM] <Ajith> yeah I guess so [10/13/2004 8:12 AM] <Srinath> yap [10/13/2004 8:12 AM] -->| Chinthaka_Away (~chinthaka@220.247.246.118) has joined #apache-axis [10/13/2004 8:12 AM] <chathura> hi all [10/13/2004 8:12 AM] <Chinthaka_Away> hi all [10/13/2004 8:12 AM] =-= Chinthaka_Away is now known as Chinthaka [10/13/2004 8:13 AM] === Chinthaka <~chinthaka@220.247.246.118> ``Eran Chinthaka'' [10/13/2004 8:13 AM] === Chinthaka: member of #apache-axis [10/13/2004 8:13 AM] === Chinthaka: attached to irc.freenode.net ``http://freenode.net/'' [10/13/2004 8:13 AM] === Chinthaka: idle for 4 seconds (on since Wednesday, October 13, 2004 8:12:21 AM) [10/13/2004 8:13 AM] --- End of WHOIS information for Chinthaka. [10/13/2004 8:13 AM] <alek> hi all [10/13/2004 8:14 AM] <Ajith> Hi all [10/13/2004 8:14 AM] <Ajith> whats on the agenda? [10/13/2004 8:15 AM] <alek> http://wiki.apache.org/ws/ChatAgenda ? [10/13/2004 8:15 AM] <Chinthaka> 1. OM design issues, [10/13/2004 8:15 AM] <Chinthaka> I updated some of the current issues [10/13/2004 8:15 AM] <Chinthaka> in wiki [10/13/2004 8:15 AM] <dims> hi all [10/13/2004 8:16 AM] [ERROR] Connection to irc://irc.freenode.net/ (irc://irc.freenode.net/) reset. [10/13/2004 8:16 AM] [INFO] Connecting to irc://irc.freenode.net/ (irc://irc.freenode.net/), attempt 1 of 5... [10/13/2004 8:17 AM] [INFO] Connecting to irc://irc.freenode.net/ (irc://irc.freenode.net/), attempt 2 of 5... [10/13/2004 8:19 AM] [INFO] Channel view for ``#apache-axis'' opened. [10/13/2004 8:19 AM] <Chinthaka> BTW, alek : shall we start our discussion with OM Design issues, while looking @ the perf results ? [10/13/2004 8:19 AM] [ERROR] Missing required parameter message. [10/13/2004 8:20 AM] =-= YOU are now known as alek_reconnected [10/13/2004 8:20 AM] <Chinthaka> (according to the Agenda :) ) [10/13/2004 8:20 AM] <alek_reconnected> i wonder if we have now all design issues? [10/13/2004 8:20 AM] <alek_reconnected> i did not see any new issues added ... [10/13/2004 8:21 AM] <Chinthaka> Hope U have seen my comments on the things u have added [10/13/2004 8:22 AM] <alek_reconnected> yes - good comments [10/13/2004 8:22 AM] <Chinthaka> :| [10/13/2004 8:22 AM] <alek_reconnected> (for reference http://wiki.apache.org/ws/FrontPage/Architecture/OM ) [10/13/2004 8:23 AM] <alek_reconnected> (see 1.2 Current issues ) [10/13/2004 8:24 AM] <Chinthaka> helloooo [10/13/2004 8:24 AM] <Ajith> yeah [10/13/2004 8:25 AM] <chathura> are we done with OM design issues if any? [10/13/2004 8:25 AM] <Ajith> I have a problem about the third issue [10/13/2004 8:25 AM] <Chinthaka> are we done with design issues, or is there any ? [10/13/2004 8:25 AM] <Ajith> making the text node just string! [10/13/2004 8:25 AM] <alek_reconnected> i think so for now [10/13/2004 8:26 AM] <Ajith> oh so we are not discussing them right now ??? [10/13/2004 8:26 AM] <Chinthaka> Ajith, for the time being, I think we can't do that in our current impls [10/13/2004 8:26 AM] <alek_reconnected> if you find new ssues please bring them to mailing list and then to wiki (when there is some consensus) [10/13/2004 8:26 AM] <Chinthaka> how do u all think abt OM performance results ? [10/13/2004 8:27 AM] <alek_reconnected> Ajith: post your view to mailing list [10/13/2004 8:27 AM] <alek_reconnected> i did not understand pefromance numbers [10/13/2004 8:27 AM] <Chinthaka> http://wiki.apache.org/ws/FrontPage/Architecture/OMPerf [10/13/2004 8:28 AM] <Chinthaka> these numbers are provided by Sosnoski's XML Bench [10/13/2004 8:28 AM] <Ajith> well it is like this [10/13/2004 8:28 AM] <Chinthaka> whats the prob u have with those results ? [10/13/2004 8:29 AM] <Ajith> I made three test classes actually with the same code to walk the tree but only the build section being different [10/13/2004 8:29 AM] <alek_reconnected> there is many zeros [10/13/2004 8:30 AM] <Chinthaka> alek : its like this [10/13/2004 8:30 AM] <alek_reconnected> results are in milliseconds? seconds? how many [10/13/2004 8:30 AM] <alek_reconnected> how many times tests were run? [10/13/2004 8:30 AM] <Chinthaka> AXTM-normal means we do not prebuilt the model [10/13/2004 8:30 AM] <Chinthaka> in the memory [10/13/2004 8:30 AM] <Ajith> building of a document in OM means just creating the doc object and thats it [10/13/2004 8:30 AM] <alek_reconnected> what is AXM-2-llm? [10/13/2004 8:31 AM] <Ajith> 50 times [10/13/2004 8:31 AM] <alek_reconnected> is it the same as AXLLM?! [10/13/2004 8:31 AM] <Chinthaka> yeah [10/13/2004 8:31 AM] <Ajith> axm llm is the OM impl with link lists [10/13/2004 8:31 AM] <alek_reconnected> what does it mean "normal"? [10/13/2004 8:31 AM] <Ajith> it has no tables inside [10/13/2004 8:32 AM] <Ajith> Normal means it is not parsing the stream in advance - effectively that is our usecase and hence the "normal" mode [10/13/2004 8:32 AM] <alek_reconnected> how minimum? walk time can be zero?! [10/13/2004 8:33 AM] <alek_reconnected> you mean headers are parsed and body is streamed? [10/13/2004 8:33 AM] <Chinthaka> thats a problem I also have [10/13/2004 8:33 AM] <Ajith> prebuilt is the object structure is built before walking [10/13/2004 8:33 AM] <Chinthaka> but Sosnoski has provided the code for JDom !!1 [10/13/2004 8:33 AM] <Ajith> well in this case we have no headers or body [10/13/2004 8:34 AM] |<-- alek has left irc.freenode.net (Read error: 110 (Connection timed out)) [10/13/2004 8:34 AM] <Ajith> even the soap XML was taken as just an XML document [10/13/2004 8:34 AM] <Chinthaka> alek : diconnected ........... [10/13/2004 8:35 AM] <Ajith> ooops [10/13/2004 8:35 AM] <Srinath> :( [10/13/2004 8:35 AM] -->| Harsha (Harsha@CPE-147-10-38-89.vic.bigpond.net.au) has joined #apache-axis [10/13/2004 8:35 AM] <alek_reconnected> i am still here :) [10/13/2004 8:36 AM] <Ajith> aah [10/13/2004 8:36 AM] <Ajith> finally :) [10/13/2004 8:36 AM] <Harsha> Hi guys [10/13/2004 8:36 AM] <Ajith> BTW hi harsha [10/13/2004 8:36 AM] <chathura> hi harsha [10/13/2004 8:36 AM] <Chinthaka> hi Harsha [10/13/2004 8:36 AM] <alek_reconnected> i have edited and hopefully labels are now consistent? http://wiki.apache.org/ws/FrontPage/Architecture/OMPerf [10/13/2004 8:37 AM] <Harsha> Asanka is here as well. [10/13/2004 8:37 AM] <Chinthaka> thats correct Alek [10/13/2004 8:38 AM] <Ajith> if not for the simple spelling mistakes you got the idea :) [10/13/2004 8:38 AM] <alek_reconnected> it looks to me that we need to remove "minimum * time" or fix how it is computed with more fine grained timer [10/13/2004 8:38 AM] <alek_reconnected> having 0 ms as time to build or walk tree is not useful informaiton ... [10/13/2004 8:39 AM] <Chinthaka> yeah [10/13/2004 8:39 AM] <Chinthaka> we were looking at the average time [10/13/2004 8:39 AM] <Ajith> hmmm [10/13/2004 8:39 AM] <Chinthaka> BTW : I corrected a mistake in the wiki, which Alek just changed [10/13/2004 8:40 AM] <alek_reconnected> in case of ATXM-nomral and AXM-2-llm there should be NA instead of zero for built time [10/13/2004 8:40 AM] <alek_reconnected> there is no build time - so it is is always zero ... [10/13/2004 8:40 AM] <alek_reconnected> now how AXM-2-llm average build-time is 1ms??? [10/13/2004 8:41 AM] <alek_reconnected> is this correct? [10/13/2004 8:42 AM] <Ajith> hmmmm [10/13/2004 8:43 AM] <Ajith> I guess this zero actually means that the time it takes is much less in order of milliseconds [10/13/2004 8:43 AM] <chathura> ajith can you pls say what invloves in the build process in the AXM 2 llm [10/13/2004 8:43 AM] <Ajith> well it is making the document object and thats it [10/13/2004 8:43 AM] <chathura> is it that you only initialise the stuff and not really build the OM am i right [10/13/2004 8:43 AM] <chathura> ok [10/13/2004 8:43 AM] <Ajith> the parser is not adavanced [10/13/2004 8:44 AM] <Ajith> i mean "advanced" [10/13/2004 8:44 AM] <alek_reconnected> i have edited again wiki page :) [10/13/2004 8:44 AM] <alek_reconnected> see now table only with avergae times [10/13/2004 8:45 AM] <Chinthaka> :) [10/13/2004 8:45 AM] <alek_reconnected> Ajith: how much initialization it does? [10/13/2004 8:46 AM] <Ajith> it only makes the document object (with no children) and thats it [10/13/2004 8:47 AM] <alek_reconnected> so it is strange that this takes 1ms?! [10/13/2004 8:47 AM] <Ajith> I will send you the code and you will get it [10/13/2004 8:47 AM] <alek_reconnected> i think we need additional test data for multiple walks - not just first walk - so it ocrresponds to case when multiple andlers are working on XML tree with headers [10/13/2004 8:47 AM] <Ajith> But yeah since you mentioned it it DOES seem unlikely [10/13/2004 8:47 AM] <alek_reconnected> Ajits: please put code in SVN [10/13/2004 8:48 AM] <Ajith> ok [10/13/2004 8:48 AM] <Chinthaka> ok, can we come to a conclusion a noew saying OM is OK ?? [10/13/2004 8:48 AM] <Ajith> I dont have SVN commit rights so I will send the code to Srinath now [10/13/2004 8:49 AM] <Chinthaka> so that we can start integrating OM to engine [10/13/2004 8:49 AM] <Chinthaka> and to see the big picture ? [10/13/2004 8:50 AM] <chathura> the numbers look pretty ok to me [10/13/2004 8:51 AM] <Chinthaka> yeah, then shall we start the discussion on WSDL Processing ? [10/13/2004 8:51 AM] <alek_reconnected> Ajith:ypu can send also patch to axis-dev and i will commit it [10/13/2004 8:52 AM] <alek_reconnected> sorry i missed it but which OM is OK? [10/13/2004 8:52 AM] <alek_reconnected> how did you come to this conclusion?! [10/13/2004 8:52 AM] <Ajith> hopefully at least the llm [10/13/2004 8:52 AM] <Chinthaka> isn't this two impls seems ok ? [10/13/2004 8:52 AM] -->| Essington (~Essington@essington.user) has joined #apache-axis [10/13/2004 8:52 AM] <alek_reconnected> what about design issues? JDOM or other API support? [10/13/2004 8:53 AM] <alek_reconnected> putting some number in wiki that are not easy to reproduce is not enough [10/13/2004 8:53 AM] <alek_reconnected> (i can not reproduce if i do not have tests source code) [10/13/2004 8:53 AM] <alek_reconnected> moreover i am concerned about what exactly tests are testing [10/13/2004 8:53 AM] <alek_reconnected> like for example how walk time correpsonds to SOAP processing model ... [10/13/2004 8:54 AM] <Chinthaka> We tested OM on Sosnpski's bench, as you suggested earlier .. [10/13/2004 8:54 AM] <chathura> the numbers are pretty comparable with jdom wrt the selected example and with differed building [10/13/2004 8:54 AM] <alek_reconnected> i can see that JDOM is faster walk time than AXM-2-llm - 2x faster ... [10/13/2004 8:55 AM] <Ajith> yep [10/13/2004 8:55 AM] <Ajith> but we did not test the memory complexity - remember?? [10/13/2004 8:55 AM] <Chinthaka> but look at the build time as well [10/13/2004 8:55 AM] <alek_reconnected> as i said number are meningless if they can not be reproduced .... you did not give information how to rerun tests, what machine was used for testing, what JDK version, ... [10/13/2004 8:56 AM] <Ajith> aah [10/13/2004 8:56 AM] <alek_reconnected> when tests multiple times do you get the same numbers? [10/13/2004 8:56 AM] <Ajith> Will you give me one day to complete the report [10/13/2004 8:56 AM] <Ajith> on this performance [10/13/2004 8:56 AM] <chathura> ok can i ask this [10/13/2004 8:57 AM] <alek_reconnected> actually JDOM seems to be 6x faster in walk time for periodic.xml - so that may make big difference when there ar emultiple SOAP handlers ... [10/13/2004 8:57 AM] <alek_reconnected> Ajith: i would prefer not ot close this until evrybody had time ot run tests ... [10/13/2004 8:58 AM] <alek_reconnected> Ajith: first thing please make test source code available [10/13/2004 8:58 AM] <Ajith> yeah I know - But I will probaly complete all the information in the wiki tomorrow [10/13/2004 8:58 AM] <alek_reconnected> ok [10/13/2004 8:58 AM] <Ajith> it is already being done [10/13/2004 8:59 AM] <Ajith> I mean Srinath is checking in the code [10/13/2004 8:59 AM] <alek_reconnected> make sure to include step-by-step instructions on who to run tests [10/13/2004 9:00 AM] <alek_reconnected> (classpath, how to integrate with sosnoski driver, etc) [10/13/2004 9:01 AM] <chathura> think i have to go now [10/13/2004 9:01 AM] <chathura> i wanted to talk about WSDL processing think i ll have to differ it or do it in da mailing list [10/13/2004 9:01 AM] <Srinath> alek see om_xmlbeanch @ scrach :) [10/13/2004 9:01 AM] <Chinthaka> I also got to go :( [10/13/2004 9:02 AM] <chathura> good night/day [10/13/2004 9:02 AM] |<-- chathura has left irc.freenode.net () [10/13/2004 9:02 AM] <Chinthaka> hopefully we can make all the things about the tests soon [10/13/2004 9:02 AM] =-= Chinthaka is now known as Chinthaka_away [10/13/2004 9:03 AM] <Ajith> I will try my best to complete it by tomorrow [10/13/2004 9:03 AM] <Harsha> We have some questions. Whom can we email them to? [10/13/2004 9:03 AM] <alek_reconnected> multiple ierations are to get meaningful results out of testing ... [10/13/2004 9:03 AM] <alek_reconnected> Harsha: i think the best place is axis-dev make sure you have [AXIS2] in subject line [10/13/2004 9:04 AM] <Harsha> alek:what is the full email address? [10/13/2004 9:04 AM] <Ajith> BTW Harsha I might be able to help you in solving some of the initial problems [10/13/2004 9:04 AM] <Srinath> ws-axis@ws.apache.org [10/13/2004 9:05 AM] <Chinthaka_away> axis-dev@ws.apache.org [10/13/2004 9:05 AM] <Srinath> sorry axis-dev [10/13/2004 9:05 AM] <Harsha> Thanks. ajith: what is your email address? [10/13/2004 9:06 AM] <Ajith> ajith.ranabahu@*.com [10/13/2004 9:06 AM] <Harsha> Thanks. Much appreciated. [10/13/2004 9:07 AM] <Ajith> BTW Alek I will try to write a small guid into running the perf tests and will work on a complete report tomorrow [10/13/2004 9:07 AM] <Ajith> Harsha : no prob [10/13/2004 9:07 AM] <Srinath> harsha:pls feel free to ask from any of in case of help :) [10/13/2004 9:08 AM] <alek_reconnected> Ajith: do not put too much work into guide - just test instructions on somebody by askiing to follow them [10/13/2004 9:08 AM] <alek_reconnected> (i can do testing but it may be more efficient f2f) [10/13/2004 9:09 AM] <Harsha> Sure. Will need the email addresses. [10/13/2004 9:09 AM] <Srinath> harsha:hemapani@*.lk [10/13/2004 9:10 AM] <Ajith> Alek : ok. But I get your point [10/13/2004 9:10 AM] <Srinath> harsha:your arddes [10/13/2004 9:10 AM] <Ajith> I mean anybody should be able to run the tests themselves and verify the results [10/13/2004 9:11 AM] <Harsha> Mine is harshap@*.au. Asanka's is jayewardena@*.com [10/13/2004 9:11 AM] <alek_reconnected> Mine is aslom@*.edu [10/13/2004 9:11 AM] <Ajith> BTW I guess the AXTM code may not be uptodate in the SVN since I have no rights to update it [10/13/2004 9:12 AM] <Srinath> I will updte it ajith;) [10/13/2004 9:12 AM] <Ajith> sure [10/13/2004 9:12 AM] <Ajith> we will do it tomorrow then [10/13/2004 9:12 AM] <Ajith> I am starving :) [10/13/2004 9:13 AM] <alek_reconnected> Ajith: send updates to axis-dev and i (or soembody else) will commit them [10/13/2004 9:13 AM] <alek_reconnected> ok [10/13/2004 9:13 AM] <alek_reconnected> i think time to finish [10/13/2004 9:13 AM] <alek_reconnected> we do not want ppl suffering ... [10/13/2004 9:13 AM] <Ajith> Alek : sure [10/13/2004 9:13 AM] <Srinath> ok :) [10/13/2004 9:13 AM] <Ajith> :) [10/13/2004 9:14 AM] <Srinath> bye all .. [10/13/2004 9:14 AM] <alek_reconnected> next week WSDL ... [10/13/2004 9:14 AM] <Harsha> Thanks and bye to all [10/13/2004 9:14 AM] <alek_reconnected> bye all [10/13/2004 9:14 AM] |<-- Harsha has left irc.freenode.net () [10/13/2004 9:14 AM] =-= YOU are now known as alek_away