Chat log

[Paul Fremantle]

[13:05] *** #apache-synapse: paulfremantle asankha ant__ dims sanjiva_ @rajith
[13:05] *** #apache-synapse was created on Mon May 08 14:27:16 2006.
[13:05] asankha: hi Paul
[13:05] paulfremantle: hi ant
[13:05] ant__: hi guys
[13:06] paulfremantle: hi *
[13:06] paulfremantle: i just sent out the vote for releas
[13:06] paulfremantle: release
[13:06] ant__: i'm playing with it right now
[13:06] paulfremantle: cool
[13:07] ant__: one thing, building from the src i have a failure in the ValidateMediatorTest
[13:07] paulfremantle: yes
[13:07] ant__: is that expected?
[13:08] ant__: org.apache.commons.httpclient.HttpException: Redirect from host www.openejb.org to www.codehaus.org is not supported
[13:08] asankha: ant you need to set up Xerces into your java/lib/endorsed
[13:08] ant__: ah ok, i thought i remembered hearing something about that
[13:08] asankha: to 'build' synapse, right now you *need* to alter ur jdk/jre lib/endorsed...
[13:09] ant__: ok
[13:09] paulfremantle: is there a way we can get maven to continue even if that test fails?
[13:09] asankha: we should be able to fix the unit tests but seems like the test plugin complicates things a bit..
[13:10] asankha: yes.. we should then add the property to ignore test errors
[13:10] paulfremantle: that would be good
[13:10] asankha: ok.. will do it
[13:11] asankha: you should also be able to just use -Dmaven.test.failure.ignore=true for the time being
[13:12] paulfremantle: sure... but its JUST that test I want to ignore :-)
[13:12] ant__: for the next release after this maybe a note in a BUILDING.txt in the src download would help
[13:12] asankha: oh.. ok..
[13:13] asankha: ok Ant.. will include that
[13:13] paulfremantle: lets put the BUILDING in now so if we rebuild for any other reasons we include it
[13:13] asankha: sure..
[13:14] paulfremantle: cool
[13:15] asankha: have created a simple http://wiki.apache.org/incubator/Synapse page to link to the updated wiki docs..
[13:15] asankha: Paul.. were you able to check with Glen whether he got a chance to test M2 RC?
[13:15] paulfremantle: nice one
[13:15] paulfremantle: i talked to him
[13:15] paulfremantle: and he said he would :-)
[13:16] paulfremantle: lets see if he's awake
[13:16] asankha: Ant.. were you able to test the samples or did you come across any issues?
[13:17] paulfremantle: no reply from Glen
[13:18] ant__: not so far, but i've only just started looking at this latest M2 download now, will continue playing with it after this chat
[13:18] asankha: ok..
[13:19] paulfremantle: well the vote is there to encourage people to look at the release :-)
[13:21] paulfremantle: so
[13:22] paulfremantle: shall we talk POST-M2?
[13:22] asankha: sure
[13:22] paulfremantle: i think there are 4/5 main things we need to do
[13:22] paulfremantle: 1) I'd love the _javascript_ and E4X support
[13:23] paulfremantle: 2) I think we need to figure out how to do pausing or continuation to support RM and call-outs
[13:23] paulfremantle: 3) A model for collecting stats.... we talked about this on earlier chats with Infravio
[13:24] paulfremantle: 4) clean up how we get at external XMLs (i.e. maybe config is in a registry)
[13:25] paulfremantle: 5) the WSDL mediation thing that Glen brought up
[13:25] paulfremantle: i'd be really happy to call a 1.0 if we had all that including RM and WS-Sec support
[13:28] asankha: Ant.. were you going to do some work related to item #1 ?
[13:29] ant__: hopefully yes. i'm already doing that in tuscany and right now tidying it up so its easy to split it for synapse as well
[13:30] asankha: great...
[13:30] asankha: i think for # 2, we should first implement a simple synchronous callout where the thread will block till the reply is received.. this would be the simple and common case..
[13:30] paulfremantle: ok.... not a bad start
[13:30] paulfremantle: but since there is also a problem with RM
[13:30] paulfremantle: we should kick around some ideas
[13:30] asankha: yes, i agree
[13:30] paulfremantle: regarding JS
[13:31] paulfremantle: I was thinking
[13:31] paulfremantle: how about us doing DOM as well as E4X
[13:31] paulfremantle: that way if there are people who don't grok E4X they can still code
[13:31] paulfremantle: what do you think Ant?
[13:32] ant__: sure should be straight forward. what about AXIOM as well?
[13:32] paulfremantle: but there isn't Axiom/JS
[13:32] paulfremantle: yet :-)
[13:33] ant__: i guess the _javascript_ mediator XML could have a parameter saying what format you want the msg in
[13:33] ant__: can't javascritp just use the Java AXIOM objects?
[13:34] ant__: if you register an OMElement object to the script environment all the methods on that object can be used by the script
[13:36] paulfremantle: i guess so....
[13:36] paulfremantle: that would be more efficient
[13:36] paulfremantle: but I really meant just use the builtin DOM apis in JS
[13:36] paulfremantle: for people who code AJAX like stuff today
[13:36] paulfremantle: using those
[13:37] ant__: don't know, i'll have to go look how that works. those DOM APIs may come from the browser environment?
[13:37] paulfremantle: yeah
[13:38] paulfremantle: i guess i assumed they are in Rhino
[13:38] paulfremantle: :-)
[13:40] asankha: shall we create a new wiki for the 1.0 feature list so that we could work against it and capture any changes from each chat to it?
[13:41] paulfremantle: good idea
[13:42] asankha: i can create the page and capture what we discuss today and the last time with glen.. else its difficult to refer back to many chat logs
[13:43] paulfremantle: +1
[13:43] paulfremantle: i've been wondering if we can use the continuation concept to deal with the RM and callout problem
[13:43] paulfremantle: its actually quite similar to the proposal I made for RM.
[13:43] paulfremantle: but we could do it a bit better
[13:44] paulfremantle: so
[13:44] paulfremantle: suppose you have a main sequence
[13:45] paulfremantle: <rules><sequence>blah blah</sequence></rules>
[13:46] paulfremantle: and in there somewhere you have something that is expected to call out asynchronously or do an ordered RM delivery
[13:46] paulfremantle: then you do
[13:46] paulfremantle: <rm><continue ref="othersequence"/></rm>
[13:46] paulfremantle: or
[13:46] paulfremantle: <call>
[13:47] paulfremantle:    .... <!-- define what to call here -->
[13:47] paulfremantle:   <continue ref="othersequence"/>
[13:47] paulfremantle: </call>
[13:48] asankha: so the reply message should execute the "othersequence"?
[13:49] paulfremantle: hmmm
[13:49] paulfremantle: not sure
[13:49] paulfremantle: maybe for callout
[13:49] paulfremantle: the current message is executed
[13:49] paulfremantle: with the reply from the callout in the context
[13:49] paulfremantle: but you might need both options
[13:52] paulfremantle: one more thing for 1.0
[13:52] paulfremantle: better fault handling
[13:52] paulfremantle: (btw... a lot of these ideas for 1.0 came from a discussion asankha and I had)
[13:53] asankha: yes.. but the difficulty i guess is how we get the reply correlated if we dont 'wait' in the current thread itself..
[13:53] paulfremantle: i think axis2 will do that for us
[13:53] paulfremantle: if we use axis2 to call out
[13:53] paulfremantle: we create a callback handler
[13:53] paulfremantle: that includes the current message and the next sequence
[13:55] asankha: ok..yes
[13:55] paulfremantle: i think thats a "good enough" model
[13:56] asankha: yes
[13:56] paulfremantle: and because its "continuations" its (1) trendy and (2) we can explain it people who understand those
[13:56] paulfremantle: :-)
[13:59] ant__: i have to go off to another meeting now
[13:59] paulfremantle: ok
[13:59] paulfremantle: thanks
[13:59] paulfremantle: see you next week
[14:00] ant__: will keep trying the M2 and vote later today/tomorrow
[14:00] paulfremantle: cool
[14:00] asankha: ok.. thanks Paul, Ant...
[14:00] paulfremantle: i'll chase glen
[14:00] paulfremantle: and post this chat

--
Paul Fremantle
VP/Technology, WSO2 and OASIS WS-RX TC Co-chair

http://bloglines.com/blog/paulfremantle
[EMAIL PROTECTED]

"Oxygenating the Web Service Platform", www.wso2.com