axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aleksander Slominski <>
Subject Re: [Axis2] Chat log 2004-10-20
Date Wed, 20 Oct 2004 14:33:55 GMT
i like inlined logs (no attachments) so here is another version with 

BTW: in future chats we should be choosing official log poster ...

summary: today discussion was about WSDL processing  and WOM (WSDL 
Object Model):
and in particular about API, data binding issues, relation to WSDL4J, 
use of XMLBeans,
metadata (get/setProperty).

we have also talked about transport and HTTP 202 support in
servlets and phased handlers and how they are supposed to work.

please make sure to visit and add topics to agenda for next week chat at:



[10/20/2004 8:00 AM] <Ias> Hi Glen
[10/20/2004 8:01 AM] =-= YOU are now known as alek
[10/20/2004 8:01 AM] <gdaniels> Hi Alek
[10/20/2004 8:01 AM] <Ajith1030> Hi all
[10/20/2004 8:02 AM] <Ias> Hi Alek,
[10/20/2004 8:02 AM] <gdaniels> Morning (evening, whatever) Ajith
[10/20/2004 8:02 AM] <Ias> Hi Ajith
[10/20/2004 8:02 AM] -->| Srinath (~Srinath@ has joined 
[10/20/2004 8:03 AM] <Ias> Hi Srinath
[10/20/2004 8:03 AM] <gdaniels> It's that time, folks.
[10/20/2004 8:03 AM] <alek> Hello
[10/20/2004 8:03 AM] <Srinath> hi all
[10/20/2004 8:03 AM] <Srinath> hi glen :)
[10/20/2004 8:04 AM] -->| Chinthaka_away (~chinthaka@ 
has joined #apache-axis
[10/20/2004 8:04 AM] <gdaniels> Hi Srinath!
[10/20/2004 8:04 AM] =-= Chinthaka_away is now known as Chinthaka
[10/20/2004 8:04 AM] <Chinthaka> hi all
[10/20/2004 8:04 AM] -->| farhaan (~Miranda@ has joined 
[10/20/2004 8:04 AM] <Ias> Hi Chinthaka
[10/20/2004 8:04 AM] <gdaniels> So... WSDL processing was apparently 
first on today's agenda?
[10/20/2004 8:04 AM] <Srinath> :)
[10/20/2004 8:05 AM] -->| dasarath (~Miranda@ has joined 
[10/20/2004 8:05 AM] <Srinath> may be we start with second chatura wsdl 
man is coming and any second
[10/20/2004 8:05 AM] <Ajith1030>
[10/20/2004 8:06 AM] <Srinath> shall I bring up handlin gbugs first
[10/20/2004 8:06 AM] <Srinath> jira for axis2
[10/20/2004 8:06 AM] <Chinthaka> thats good Srianth
[10/20/2004 8:06 AM] <Srinath> gbug=bug
[10/20/2004 8:06 AM] <Chinthaka> Srinath :)
[10/20/2004 8:06 AM] <gdaniels> yes, should be pretty easy to set up
[10/20/2004 8:06 AM] <farhaan> okay, so what do you propose? I would 
prefer to have a a new jIRA project created
[10/20/2004 8:06 AM] <gdaniels> What did you want to discuss about it?
[10/20/2004 8:06 AM] <gdaniels> farhaan: yup
[10/20/2004 8:06 AM] <Srinath> should we set up a jira project to axis2
[10/20/2004 8:07 AM] <Srinath> or do it inside axis jira
[10/20/2004 8:07 AM] <gdaniels> +1 new
[10/20/2004 8:07 AM] <Srinath> +1 new
[10/20/2004 8:08 AM] <Srinath> ok let us catch dims for that i belive
[10/20/2004 8:08 AM] <Ajith1030> I woluld also say new
[10/20/2004 8:08 AM] <gdaniels> OK
[10/20/2004 8:08 AM] <alek> +1 new
[10/20/2004 8:09 AM] <Srinath> next the OM how to proceed from where we are
[10/20/2004 8:09 AM] <Chinthaka> wait
[10/20/2004 8:09 AM] <Chinthaka> shall we discuss this wsdl first
[10/20/2004 8:10 AM] <Chinthaka> its accrued from last chat :)
[10/20/2004 8:10 AM] -->| chathura (~chathura@ has 
joined #apache-axis
[10/20/2004 8:10 AM] <gdaniels> We're waiting for Chathura
[10/20/2004 8:10 AM] <gdaniels> ah
[10/20/2004 8:10 AM] <Chinthaka> he came just now
[10/20/2004 8:10 AM] <gdaniels> speak of the devil
[10/20/2004 8:10 AM] <chathura> hi sorry guy
[10/20/2004 8:10 AM] <Srinath> Chat is in :)
[10/20/2004 8:10 AM] <Chinthaka> :D
[10/20/2004 8:10 AM] <gdaniels> no problemo
[10/20/2004 8:10 AM] <Srinath> then to wsdli belive
[10/20/2004 8:10 AM] <chathura> +1:)
[10/20/2004 8:10 AM] <gdaniels> So OK, WSDL first then OM
[10/20/2004 8:11 AM] <Ajith1030> yeah Chathura :  how is your progress
[10/20/2004 8:11 AM] <chathura> ok will tell you what i have in mind
[10/20/2004 8:11 AM] <chathura> ok i have a prototype impl of a possible WOM
[10/20/2004 8:11 AM] <Chinthaka> all of us are staring at him and he is 
bit nervous :)
[10/20/2004 8:11 AM] <chathura> WOM = WSDL object Model
[10/20/2004 8:12 AM] <chathura>
[10/20/2004 8:12 AM] <chathura> its the architecture i have in mind 
which is pretty abstract(of course)
[10/20/2004 8:13 AM] <chathura> THis WOM of mine will be based on the 
WSDL 2.0 component model
[10/20/2004 8:14 AM] <gdaniels> Are we expecting the single WOM model to 
deal with both WSDL 2.0 and 1.1?
[10/20/2004 8:14 AM] <chathura> but there are sufficient provisions that 
we can keep the
[10/20/2004 8:14 AM] <chathura> wsdl 1.1:)
[10/20/2004 8:15 AM] -->| Srinath_ (~Srinath@ has joined 
[10/20/2004 8:15 AM] <Srinath_> hi are you seening this
[10/20/2004 8:15 AM] <Chinthaka> yeah, y not :)
[10/20/2004 8:15 AM] <Chinthaka> seems you guys having some prob
[10/20/2004 8:15 AM] <gdaniels> I notice you're storing things in the 
current WOM proto as LinkedLists... wouldn't Maps make more sense, 
indexed by QNames/NCNames as appropriate?
[10/20/2004 8:16 AM] <Srinath_> i think the net goes down and come back :D
[10/20/2004 8:16 AM] <Srinath_> :(:(
[10/20/2004 8:16 AM] <gdaniels> Srinath: yes, we see your clone :)
[10/20/2004 8:16 AM] <Chinthaka> :)
[10/20/2004 8:16 AM] <Srinath_> :D
[10/20/2004 8:16 AM] -->| Ajith (~Miranda@ has joined 
[10/20/2004 8:17 AM] <Srinath_> ajith can you address the glen Q about OM?
[10/20/2004 8:17 AM] <chathura> glen:sounds good
[10/20/2004 8:17 AM] <Ajith> sorry guys
[10/20/2004 8:17 AM] <Ajith> what was the question??
[10/20/2004 8:17 AM] <Chinthaka> Did Glen asked anything abt OM ?
[10/20/2004 8:17 AM] <Ajith> I missed a whole lot
[10/20/2004 8:18 AM] <gdaniels> <gdaniels> I notice you're storing 
things in the current WOM proto as LinkedLists... wouldn't Maps make 
more sense, indexed by QNames/NCNames as appropriate?
[10/20/2004 8:18 AM] -->| dasarath1318 (~Miranda@ has 
joined #apache-axis
[10/20/2004 8:18 AM] <gdaniels> I think that was for Chathura, though - 
he wrote the current WOM, yes?
[10/20/2004 8:18 AM] <chathura> well one more thig that i forgot to mension
[10/20/2004 8:18 AM] <Srinath_> sorry it is WOM :)
[10/20/2004 8:18 AM] <chathura> glen : didnt get you :(
[10/20/2004 8:19 AM] <gdaniels> In, for instance - 
you have the interfaces stored in a LinkedList.
[10/20/2004 8:20 AM] <chathura> yes yes
[10/20/2004 8:20 AM] <chathura> the thing i wanted to tell you was
[10/20/2004 8:20 AM] <chathura> the need for external xml binding mechanism
[10/20/2004 8:20 AM] <gdaniels> i.e. an external reader/writer?
[10/20/2004 8:20 AM] <chathura> more or less the direcion is toward xml 
beans i guess
[10/20/2004 8:21 AM] -->| farhaan5843 (~Miranda@ has 
joined #apache-axis
[10/20/2004 8:21 AM] <chathura> no i was talking about the schema 
parsing and the data representation of the WOM
[10/20/2004 8:22 AM] <chathura> glen : yes your one plus and the data 
representation of the WOM
[10/20/2004 8:22 AM] <gdaniels> ok, are you suggesting using XMLBeans on 
the WSDL schema to generate the classes?
[10/20/2004 8:22 AM] <Ias> Actually Beehive WSM does that.
[10/20/2004 8:23 AM] <chathura> think the Ajith is having some problems 
with the chat
[10/20/2004 8:23 AM] <gdaniels> hm.  I would think that would tie it too 
closely to a particular version....
[10/20/2004 8:23 AM] <Ias> A similar case happened to EWS.
[10/20/2004 8:24 AM] <Ias> Once we used JAXB, and Geronimo folks wanted 
us to change to XMLBeans.
[10/20/2004 8:24 AM] -->| Ajith_rana (~Miranda@ has joined 
[10/20/2004 8:24 AM] <Ajith_rana> guys  we are having probs here at UOC
[10/20/2004 8:24 AM] <chathura> i think the code gen part should be 
clearly seperated from this WOM of mine
[10/20/2004 8:24 AM] <Ajith_rana> so our connection seems to break always
[10/20/2004 8:25 AM] <Ias> Now it seems common that we need to make some 
abstract layer for supporting various binding solutions.
[10/20/2004 8:25 AM] <gdaniels> OK, so you're really suggesting using 
the codegen to read and write the XML?  That means that you'll have to 
write/read the XMLBeans objects to/from the WOM objects...
[10/20/2004 8:25 AM] <chathura> so glen the code generation will be 
based on the WOM inside of which we may use xml beans and it will 
eventually help us
[10/20/2004 8:26 AM] -->| Srinath__ (~Srinath@ has joined 
[10/20/2004 8:26 AM] <gdaniels> I'm not sure there's a big enough payoff 
there to make XMLBeans worth it over a simpler solution, but I'll hold 
judgement until I see it.
[10/20/2004 8:26 AM] <Srinath__> well am keep cloning :(
[10/20/2004 8:26 AM] <gdaniels> I'm not sure there's NOT a big enough 
payoff either. :)
[10/20/2004 8:27 AM] <gdaniels> If we have 10 Srinaths, we'll get this 
thing done a lot faster!
[10/20/2004 8:27 AM] <Srinath__> oh :D
[10/20/2004 8:27 AM] <Ias> :-)
[10/20/2004 8:27 AM] <Ajith_rana> :D
[10/20/2004 8:28 AM] <chathura> glen:i didnt get what you really mean
[10/20/2004 8:28 AM] <alek> :-)
[10/20/2004 8:28 AM] <chathura> we ll pretty much depend on the xml 
beans to d  the schema parsing
[10/20/2004 8:28 AM] <chathura> we need a schema parser anyway
[10/20/2004 8:28 AM] <alek> so WOM will have one common API and support 
both WSDL2 and WSDL1.1?

[10/20/2004 8:29 AM] <gdaniels> chathura: I'm saying that I'm a little 
concerned that the XMLBeans model generated from WSDL will duplicate an 
awful lot of what's going to be in the abstract WOM, and that it might 
be faster/easier to have a specific reader/writer class that does 
WOM<->XML directly.
[10/20/2004 8:29 AM] <Srinath__> we got to decide what to use 4 xml 
schmema interface + parsing
[10/20/2004 8:29 AM] <Ajith_rana> yeah we can the schema support from XB!
[10/20/2004 8:29 AM] |<-- Srinath has left (Read error: 
113 (No route to host))
[10/20/2004 8:29 AM] <chathura> alek: yes
[10/20/2004 8:30 AM] |<-- Ajith1030 has left (Read 
error: 113 (No route to host))
[10/20/2004 8:30 AM] |<-- dasarath has left (Read 
error: 113 (No route to host))
[10/20/2004 8:30 AM] |<-- farhaan has left (Read error: 
113 (No route to host))
[10/20/2004 8:30 AM] <chathura> glen: my  guess is that would lead us to 
the place we are at in axis 1.1
[10/20/2004 8:30 AM] <gdaniels> We wanted to have pluggable data binding 
frameworks, right?  If that's the case then we want to think about what 
we need from the abstract part...
[10/20/2004 8:31 AM] <chathura> i agree
[10/20/2004 8:31 AM] |<-- FR^2 has left (Read error: 
110 (Connection timed out))
[10/20/2004 8:31 AM] <gdaniels> chathura: In Axis 1.1 we use WSDL4J
[10/20/2004 8:32 AM] <gdaniels> WSDL4J does the WSDL writing/reading for us
[10/20/2004 8:32 AM] <chathura> and we do our own schema parsing
[10/20/2004 8:32 AM] <chathura> yah
[10/20/2004 8:32 AM] <alek> glen: by pluggable do you mean that WSDL 
code generator can support many XML<->Java classes mappings?
[10/20/2004 8:32 AM] <Ias> What's going on with WSDL4J 2?
[10/20/2004 8:32 AM] <gdaniels> Ias: Nothing yet
[10/20/2004 8:32 AM] -->| FR^2 
( has joined #apache-axis
[10/20/2004 8:32 AM] <Ias> Glen: then when?
[10/20/2004 8:32 AM] <gdaniels> Ias: We discussed contributing our 
version to JCP when we were in Sri Lanka...
[10/20/2004 8:33 AM] <Ias> Glen: JCP? not Apache?
[10/20/2004 8:33 AM] <gdaniels> alek: Yes, but probably not 
mix-and-match, just one at a time.
[10/20/2004 8:33 AM] <gdaniels> Ias: "our" == Apache's
[10/20/2004 8:33 AM] <chathura> glen: do you not think that schema 
parsing is a good reason for us to think of pluggable data binding
[10/20/2004 8:34 AM] -->| dasarath (~Miranda@ has joined 
[10/20/2004 8:34 AM] <gdaniels> chathura: Oh - you're talking about 
XMLBeans for parsing the schemas
[10/20/2004 8:34 AM] <gdaniels> INSIDE the WSDL
[10/20/2004 8:34 AM] <gdaniels> not the schema for WSDL itself
[10/20/2004 8:34 AM] <chathura> not really
[10/20/2004 8:34 AM] <gdaniels> I misunderstood, I think
[10/20/2004 8:34 AM] <chathura> its like this
[10/20/2004 8:35 AM] <Ias> Glen: I see. What I remember about that is 
WSDL4J 2 code base will come from IBM. Therefore, it is not ready yet.
[10/20/2004 8:35 AM] <chathura> xml beans (or the data binding ) will be 
used for two things i belive
[10/20/2004 8:35 AM] <gdaniels> Ias: No, WSDL4J2 code base isn't really 
being worked on yet IIRC.  We suggested contributing our thing as the 
basis for WSDL4J2 work when we're done.
[10/20/2004 8:36 AM] <gdaniels> I might be remembering wrong, though.
[10/20/2004 8:36 AM] <chathura> one is to parse the schema component 
that is embedded in the WSDL itself
[10/20/2004 8:36 AM] <chathura> THis will go into the WSDLOM Builder 
inthe diagram in
[10/20/2004 8:36 AM] <gdaniels> chathura: +1
[10/20/2004 8:36 AM] <chathura> :)
[10/20/2004 8:37 AM] <Ias> +1 from me too.
[10/20/2004 8:37 AM] |<-- Srinath_ has left (Read 
error: 113 (No route to host))
[10/20/2004 8:37 AM] <chathura> that was the first one by the way:)
[10/20/2004 8:37 AM] <gdaniels> chathura: and thing number two?
[10/20/2004 8:37 AM] -->| Srinath (~Srinath@ has joined 
[10/20/2004 8:37 AM] <Ias> Srinath comes and goes :-)
[10/20/2004 8:37 AM] <Srinath> :(
[10/20/2004 8:37 AM] <Srinath> at least am keep cloning
[10/20/2004 8:38 AM] <Srinath> :D
[10/20/2004 8:38 AM] <chathura> No 2 is that we will use the xml beans 
object model to represent the schema component inside the WOM
[10/20/2004 8:38 AM] <Chinthaka> :)
[10/20/2004 8:38 AM] |<-- Ajith has left (Read error: 
113 (No route to host))
[10/20/2004 8:38 AM] <chathura> its not in da code as yet
[10/20/2004 8:39 AM] |<-- dasarath1318 has left (Read 
error: 113 (No route to host))
[10/20/2004 8:40 AM] <gdaniels> OK
[10/20/2004 8:40 AM] <chathura> so WOM will have the wsdl component 
model plus references to a object model provided to us by a data binding 
[10/20/2004 8:40 AM] <gdaniels> sounds good
[10/20/2004 8:40 AM] <chathura> :)
[10/20/2004 8:40 AM] <gdaniels> WOM interfaces need to make it easy to 
find components from various places by name
[10/20/2004 8:41 AM] <gdaniels> WOM also needs to be able to hang 
arbitrary data off each component
[10/20/2004 8:41 AM] <chathura> glen: i agree with you we ll adopt a 
better serch mechanism
[10/20/2004 8:41 AM] <gdaniels> (so we can attach handler chains, etc)
[10/20/2004 8:41 AM] <chathura> explain a bit more pls
[10/20/2004 8:42 AM] <gdaniels> each component should have a 
[10/20/2004 8:42 AM] <gdaniels> that lets us attach handler chains to 
the Input/Output WSDL objects, as discussed in SL
[10/20/2004 8:42 AM] <gdaniels> (operation/service too)
[10/20/2004 8:42 AM] <gdaniels> plus attach any other metadata we might need
[10/20/2004 8:43 AM] <chathura> so there will  be some thing like a 
[10/20/2004 8:43 AM] <chathura> ?
[10/20/2004 8:43 AM] <gdaniels> more like a WSDLComponent base class
[10/20/2004 8:44 AM] <gdaniels> which implements the property-storing stuff
[10/20/2004 8:45 AM] <alek> what is Property? named Object?
[10/20/2004 8:45 AM] <gdaniels> right
[10/20/2004 8:45 AM] <chathura> i was thinking some think like 
MessageContext like thing per component yah the impl would converge us 
to what you said
[10/20/2004 8:45 AM] -->| sanjiva (~sanjiva@ has joined 
[10/20/2004 8:45 AM] <gdaniels> Chain chain = 
[10/20/2004 8:46 AM] <sanjiva> hi Guys
[10/20/2004 8:46 AM] <sanjiva> sorry I'm rather late ..
[10/20/2004 8:46 AM] <gdaniels> yo Sanjiva
[10/20/2004 8:46 AM] <chathura> hi sir
[10/20/2004 8:46 AM] <gdaniels> no worries, but we assigned you a whole 
lot of code to write while we were waiting. :D
[10/20/2004 8:46 AM] <sanjiva> what's the topic today?
[10/20/2004 8:46 AM] <sanjiva> ah cool :)
[10/20/2004 8:46 AM] <Chinthaka> :)
[10/20/2004 8:46 AM] <gdaniels> we're talking WSDL right now
[10/20/2004 8:47 AM] <gdaniels> coming to a close on it for today, I think
[10/20/2004 8:47 AM] <gdaniels> chathura's working on the object model, 
in early stages
[10/20/2004 8:47 AM] <Ias> Hi Sanjiva
[10/20/2004 8:47 AM] <alek> hi sanjiva
[10/20/2004 8:47 AM] <gdaniels> we discussed the general idea of 
plugging in a databinding/schema framework to do the <types> work, but 
haven't talked specifics yet.
[10/20/2004 8:47 AM] <gdaniels> XMLBeans first, pluggability after, I think.
[10/20/2004 8:48 AM] <sanjiva> ok sounds good ... but we will have the 
option of exposing OM directly to the app if they want it so right?
[10/20/2004 8:48 AM] <gdaniels> right
[10/20/2004 8:48 AM] <sanjiva> I mean basically without any type mapping
[10/20/2004 8:48 AM] <gdaniels> yes
[10/20/2004 8:48 AM] <sanjiva> OK
[10/20/2004 8:48 AM] <gdaniels> will we do schema validation in that case?
[10/20/2004 8:49 AM] <sanjiva> good question .. not yet I guess but 
certainly going forward that would be a good thing to do
[10/20/2004 8:49 AM] <gdaniels> +1 not yet, and also to aim for that later
[10/20/2004 8:49 AM] <sanjiva> bit of a challenge to do schema 
validation while doing deferred tree building with a pull parser but not 
impossible for sure
[10/20/2004 8:49 AM] |<-- Srinath__ has left (Read 
error: 242 (No route to host))
[10/20/2004 8:50 AM] <Chinthaka> deferred tree building in WOM ??
[10/20/2004 8:50 AM] <gdaniels> right - technically can't validate until 
the whole infoset is there, but you can do stepwise validation with 
faults happening whenever the bad elements are encountered...
[10/20/2004 8:50 AM] <gdaniels> Chinthaka: no, OM
[10/20/2004 8:50 AM] |<-- Ajith_rana has left (Read 
error: 113 (No route to host))
[10/20/2004 8:50 AM] |<-- farhaan5843 has left (Read 
error: 113 (No route to host))
[10/20/2004 8:50 AM] <gdaniels> OK, speaking of OM.... should we move on 
to that?
[10/20/2004 8:51 AM] <Chinthaka> are we talking abt OM ...... ??????
[10/20/2004 8:51 AM] <Chinthaka> or WOM .... ??? or both ??????
[10/20/2004 8:51 AM] <gdaniels> Chinthaka: Yes, Sanjiva and I were 
talking about using OM directly instead of data binding when desired.
[10/20/2004 8:52 AM] <alek> it seems there are soem connectivity problems
[10/20/2004 8:52 AM] <sanjiva> Chinthaka: about givinmg the app 
programmer a choice of programming models: type mapped into a java type 
or as an XML infoset (aka OM)
[10/20/2004 8:52 AM] <alek> maybe we can move OM to next week and rehash 
issues over mailing list?
[10/20/2004 8:53 AM] <gdaniels> Sure, we can do that.
[10/20/2004 8:53 AM] <gdaniels> Srinath, are you still there?
[10/20/2004 8:53 AM] <Chinthaka> alek : thats good
[10/20/2004 8:53 AM] <Chinthaka> its lightening here :(
[10/20/2004 8:53 AM] <gdaniels> looks like no :(
[10/20/2004 8:53 AM] <chathura> think the guys at uoc having problem
[10/20/2004 8:53 AM] <sanjiva> you mean to say he got hit??
[10/20/2004 8:54 AM] <sanjiva> ah ok ;-)
[10/20/2004 8:54 AM] <chathura> ;)
[10/20/2004 8:54 AM] <gdaniels> lightning == bad for the wires, not the 
people! :)
[10/20/2004 8:54 AM] <gdaniels> So I checked in an engine prototype
[10/20/2004 8:54 AM] -->| smeder 
( has joined #apache-axis
[10/20/2004 8:54 AM] <alek> unless you have really really really bad luckk
[10/20/2004 8:54 AM] <Chinthaka> people are infront of things powered by 
wires ;)
[10/20/2004 8:55 AM] <gdaniels> Will do some more work on it, but it 
shows some of the stuff we talked about in early form.
[10/20/2004 8:55 AM] <gdaniels> Hi Sam
[10/20/2004 8:55 AM] <Chinthaka> hi sam
[10/20/2004 8:55 AM] <Ias> Hi Sam
[10/20/2004 8:55 AM] <smeder> hi all
[10/20/2004 8:55 AM] <chathura> HI SAM
[10/20/2004 8:55 AM] <sanjiva> Glen: saw it .. I'll look at it too (have 
to look at srinath's code too)
[10/20/2004 8:55 AM] <sanjiva> hi Sam!
[10/20/2004 8:56 AM] <Chinthaka> a small question ... can some one tell 
me abt
[10/20/2004 8:56 AM] <Chinthaka> what Phased handlers means ... for 
[10/20/2004 8:56 AM] <gdaniels> the idea that there are well-defined 
"phases" with names that represent particular ordered parts of your 
handler chain
[10/20/2004 8:56 AM] |<-- Srinath has left (Read error: 
113 (No route to host))
[10/20/2004 8:57 AM] <alek> hi Sam
[10/20/2004 8:57 AM] <gdaniels> phases get used by Modules when they're 
deploying handlers.
[10/20/2004 8:57 AM] |<-- dasarath has left (Read 
error: 113 (No route to host))
[10/20/2004 8:57 AM] <gdaniels> So you can say stuff like "deploy this 
handler as the first one in the dispatch phase", etc
[10/20/2004 8:57 AM] <Chinthaka> oki
[10/20/2004 8:58 AM] <gdaniels> a Phase as I built it has pre and post 
conditions which get checked during execution
[10/20/2004 8:58 AM] <alek> what would be typical phases? ike security, 
reliable messaging, ... ?
[10/20/2004 8:58 AM] <gdaniels> alek: eventually, yes
[10/20/2004 8:58 AM] <gdaniels> we'll start simple with just transport, 
global, dispatch....
[10/20/2004 8:59 AM] <Chinthaka> did u see the protoype we created for 
Engine ?
[10/20/2004 8:59 AM] <gdaniels> I've got it but haven't looked at it yet 
- plan to today
[10/20/2004 9:00 AM] <chathura> i have looked at the transport sender thing
[10/20/2004 9:01 AM] <chathura> few issues in my mind
[10/20/2004 9:01 AM] <gdaniels> which transport sender thing?
[10/20/2004 9:02 AM] <chathura> The transport sender thing that we 
talked about in the sumit
[10/20/2004 9:02 AM] <chathura> the one who will od the transport 
realated stuff
[10/20/2004 9:03 AM] <chathura> eg: sending 202 at the end of the in 
phase if necessary
[10/20/2004 9:03 AM] <chathura> one who will be the last handler in the out
[10/20/2004 9:04 AM] <chathura> basically the guy who will look up 
whether ther are addressing things to be clarified before sending the out
[10/20/2004 9:04 AM] <gdaniels> I implemented that one of the ways we 
talked about, by having the servlet (HTTPListener) send the 202 if it 
gets control back but does not find a particular RESPONSE_SENT property 
in the MC
[10/20/2004 9:05 AM] <chathura> yes yes thats what i ws talking
[10/20/2004 9:05 AM] <gdaniels> So if someone used the HTTPResponder 
(the "sender") to send a response or a fault, the property gets set and 
the HTTPListener does nothing when it gets control back.  Otherwise it 
sends a 202.
[10/20/2004 9:06 AM] <chathura> so if i am to tell you my question
[10/20/2004 9:06 AM] <gdaniels> The only problem with that approach is 
that the 202 gets sent AFTER all the processing has happened.  It might 
also be nice to put in a way to have that happen earlier.
[10/20/2004 9:07 AM] <chathura> yah in mind i thought this transport 
sender gets called the responder
[10/20/2004 9:07 AM] <chathura> befor out after in
[10/20/2004 9:07 AM] <sanjiva> glen: that would require another thread 
.. otherwise the servlet rules will be broken (?)
[10/20/2004 9:07 AM] <gdaniels> sanjiva: no, I don't think so
[10/20/2004 9:07 AM] <gdaniels> Anyone can write and flush the servlet 
[10/20/2004 9:08 AM] <gdaniels> We just need to know when to do it.
[10/20/2004 9:08 AM] <sanjiva> hmmm .. even before the return from 
[10/20/2004 9:09 AM] <gdaniels> There's some kind of "switch responders" 
hook that would perform activity like sending the 202 and flushing.
[10/20/2004 9:09 AM] <gdaniels> Yes, I'm pretty sure.
[10/20/2004 9:09 AM] <gdaniels> Not positive though.
[10/20/2004 9:09 AM] <chathura> what will happen in cases like 
ws-eventing where the same message is sent out to many listeners
[10/20/2004 9:09 AM] <chathura> two questions
[10/20/2004 9:10 AM] <gdaniels> chathura: ? Not sure what you're getting 
at with eventing.
[10/20/2004 9:10 AM] <sanjiva> so you mean if I do this in in doPost: { 
.. code to send reponse, flush; sleep (100000); return } then the client 
will receive the 200 OK response before the sleep expires?
[10/20/2004 9:10 AM] <chathura> does the current addressing allow us to 
have many reply tos
[10/20/2004 9:10 AM] <gdaniels> sanjiva: I think so, yes.
[10/20/2004 9:10 AM] <gdaniels> it probably doesn't close down the 
socket, but you get the data anyway.
[10/20/2004 9:11 AM] <alek> maybe doPost can examine handler chain 
step-wise and if any step delcares that no response will be sent (such 
as for one-way messaging) then 202 can be sent and chain processing can 
[10/20/2004 9:12 AM] <sanjiva> chathura: the servlet stuff comes in for 
a request-response type pattern only ... in the case of eventing its 
just sending out multiple messsages only right?
[10/20/2004 9:12 AM] <gdaniels> alek: not sure you need to examine the 
whole chain for that, just determine the MEP
[10/20/2004 9:12 AM] <sanjiva> I mean the implementation of ws-eventing 
is going to use a sender multiple times - it will not involve a servlet 
at all
[10/20/2004 9:13 AM] <chathura> sir i was wondering whether we can burn 
that part into the transport sender itself
[10/20/2004 9:13 AM] <chathura> after all it ll be the same msg right
[10/20/2004 9:13 AM] <gdaniels> burn the idea of sending to multiple 
destinations into the sender itself?
[10/20/2004 9:14 AM] <gdaniels> You certainly could write a sender like 
that, but I don't think we need to do anything architecturally to 
support it.
[10/20/2004 9:14 AM] <chathura> meaning if the eventing requires sending 
same msg to 10 listenenrs it ll be a single out but our transport sender 
will will send it out to all
[10/20/2004 9:15 AM] |<-- FR^2 has left (Read error: 
110 (Connection timed out))
[10/20/2004 9:15 AM] <gdaniels> but it might need to be encrypted 
differently for each listener
[10/20/2004 9:15 AM] -->| FR^2 (~FRQuadrat@ has joined 
[10/20/2004 9:15 AM] <gdaniels> so you really need to call the whole 
outgoing chain for each message to enable that sort of thing
[10/20/2004 9:16 AM] <chathura> ohh ok
[10/20/2004 9:16 AM] <sanjiva> ah ok .. in some sense a multiplexing 
sender instead of a point-to-point sender. interesting ..
[10/20/2004 9:16 AM] <gdaniels> should be cheap in most cases
[10/20/2004 9:16 AM] <chathura> :)
[10/20/2004 9:17 AM] <alek> well that may got more complicate with WSRM 
as sender may need to retry sending so separate scheduler (or thread) is 
[10/20/2004 9:17 AM] <sanjiva> but we definitely don't need the sender 
to implement ws-rm .. that can be layered on top
[10/20/2004 9:17 AM] <sanjiva> never mind that's what you said :)
[10/20/2004 9:18 AM] <gdaniels> OK, shall we wrap it up for today?  
Everyone go look at everyone else's code. :)
[10/20/2004 9:19 AM] <chathura> think i need to go now today is our 
graduation ball at the university
[10/20/2004 9:19 AM] <chathura> ok
[10/20/2004 9:19 AM] <gdaniels> I'm psyched to play with the OM protos!
[10/20/2004 9:19 AM] <sanjiva> sounds good :)
[10/20/2004 9:19 AM] <sanjiva> congrats on the ball guys! have a ball! ;-)
[10/20/2004 9:19 AM] <gdaniels> chathura: have fun!
[10/20/2004 9:19 AM] <gdaniels> hehehe
[10/20/2004 9:19 AM] <sanjiva> bye ..
[10/20/2004 9:19 AM] <chathura> thanx
[10/20/2004 9:19 AM] <chathura> bye
[10/20/2004 9:19 AM] <--| chathura has left #apache-axis
[10/20/2004 9:19 AM] <--| smeder has left #apache-axis
[10/20/2004 9:19 AM] <gdaniels> bye4now folks... see you all next week
[10/20/2004 9:19 AM] <gdaniels> I'll send the log
[10/20/2004 9:19 AM] <Ias> Bye+
[10/20/2004 9:20 AM] <alek> bye
[10/20/2004 9:20 AM] <alek> i will send log too

The best way to predict the future is to invent it - Alan Kay

View raw message