Differences between revisions 4 and 5
Revision 4 as of 2005-12-10 02:33:57
Size: 15935
Editor: 216
Revision 5 as of 2009-09-20 22:48:55
Size: 15935
Editor: localhost
Comment: converted to 1.6 markup
No differences found!

Chat log

  • [10:27] Ajith: . Have a interface for the repsitory [10:27] Ajith: so that a repo does not depend on the FS

    [10:27] Ajith: FS -> File System [10:39] dims pokes the team

    [10:40] dims: are u guys awake? :) [10:40] Chinthaka: yes [10:40] Ajith: yep [10:40] Ajith: Glen's on a call

    [10:40] Chinthaka: another call :) [10:40] dims: busy guy

    [10:41] dims: :) [10:43] dims: and where is the boss? [10:44] Chinthaka: boss ?

    [10:45] Deepal: who is the boss :) ? [10:45] dims mutters...sanjiva? [10:46] Ajith: Boss is on the way I guess [10:46] Chinthaka: BTW, we started to think abt a file system abstraction, which we wanted to do some time back too [10:46] dims: ok [10:46] Chinthaka: that is a big ISSUE and will take some time .. So we will leave it behind, after the discussion [10:47] dims: yep [10:47] Chinthaka: we are fixing whatever spelling mistakes, as and when we find them [10:47] dims: lots of things to do before we take that one on. [10:47] dims: +1 [10:48] Chinthaka: BTW, when are you coming here ? Tomoro afternoon ? [10:48] Chinthaka: time ? [10:49] dims: yes [10:49] dims: around 1:00 PM [10:51] dims: i'll take the shuttle and get to the hotel and try to find u guys [10:51] dims: when is paul reaching? [10:56] Chinthaka: I think Paul and Sanjiva are coming together by 2.00 p.m.

    [11:00] Chinthaka: I think we are converging for a design on AxisConfig abstraction here [11:01] Chinthaka: so that, we do not depend on file system or anything

    [11:01] Chinthaka: Glen : Would you like to explain that here, please .. :) And we can move on with debugging ... [11:16] FR^2 has joined [11:29] Chinthaka has disconnected: Read error: 113 (No route to host) [11:30] Chinthaka has joined [11:31] Deepal: I think we need to discuss how to miggrete from Axis 1.x [11:32] Deepal: we need to provide backword compatibility [11:32] Chinthaka: hey guys [11:32] Chinthaka: sorry I went up for a Coke [11:32] Chinthaka: Paul ?? [11:32] Chinthaka: where are you ? [11:36] gdaniels has disconnected: Read error: 110 (Connection timed out) [11:39] Glen-away has joined [11:39] Glen-away is now known as gdaniels [11:39] gdaniels: OK, my summary...

    [11:40] gdaniels: We've agreed that we want a "discovery-like" way of getting AxisConfigurations

    [11:41] gdaniels: We want a ConfigurationFactory (or whatever we call it) that needs a couple of kinds of information - 1) the class name of the actual factory (i.e. FileConfigurationFactory, DBConfigurationFactory), and 2) a bag of properties which will be used by the instance factory to do work (i.e. filename, etc)

    [11:42] gdaniels: That thing will give us an AxisConfiguration. It will be able to default by using system properties, home directory configuration files, etc... (i.e. AXIS_CONFIG=org.apache.axis2.config.FileConfigurationFactory)

    [11:43] gdaniels: So there are now two ways we can look at AxisConfiguration (WAKE UP EVERYONE THIS IS WHERE WE ASK YOU A QUESTION)..

    [11:43] gdaniels: 1) AxisConfiguration is read-only. It gets its data from the filesystem or DB or memory, but if you want to *write* to it, you need another object (maybe the instance of the configuration factory)

    [11:44] gdaniels: 2) AxisConfiguration is read-write, and has "deployService()" type methods on it. These would do the right thing in terms of writing to the filesystem/DB/backing-store. [11:44] gdaniels: Assuming I got that right, we're wondering what people think. [11:45] gdaniels: so... is anyone there aside from us warm sunny San Diego folks? [11:45] dims: +1 [11:45] gdaniels: So dims, do you like (1) or (2), or do you need more info to clarify?

    [11:46] Chinthaka: warm ? I prefer it over the snowy Boston ;)

    [11:46] dims: So we could have a spring based AxisConfiguration [11:46] gdaniels: yes

    [11:46] gdaniels: the real question is do we want AxisConfiguration to be read/write, or just read [11:46] dims: let's start with 1 [11:47] dims: read-only [11:47] Chinthaka: +1 for read only [11:47] gdaniels: Chinthaka and Glen like (2)

    [11:47] gdaniels: oops, maybe I was wrong :) [11:47] Chinthaka: at least for now ... [11:47] dims: write is specific to the configuration provider [11:47] gdaniels: but we do need a generic interface for writing [11:48] Chinthaka: I'm talking abt the current plan and then to support read only later [11:48] dims: let's start with read only and then revisit if we can live w/o a write

    [11:48] Ajith: I like 1 since for me AxisConfig is only a infostore for ne [11:48] Ajith: me [11:48] dims: BUT the problem is for example synapse wants to add dyanmic services on the fly?

    [11:49] gdaniels: my preference is to make AxisConfiguration read/write, which makes building in-memory versions,etc, very easy [11:49] gdaniels: dims: +1 [11:49] Ajith: so basically even if you have several implementations the diff would only be in the read write code [11:49] dims: argh!

    [11:49] gdaniels: then the individual instances can implement write operations in their own way, even if they want to throw UnsupportedOperation [11:49] Deepal: my idea is there shoud be write capability some where [11:49] Deepal: it might be in AxisOCnfig or Context

    [11:49] gdaniels: Deepal: I sure hope so :) [11:49] Ajith: so we can push the read write code to a seperate builder/creator componet

    [11:50] dims: another trouble is how to notify the engine that the AxisConfig got updated [11:50] dims: (someone drops a aar in the FS) [11:50] Ajith: BTW my guess is 90% case we will read from the config rather than writing from it [11:50] dblevins has joined [11:50] Chinthaka: sorry, confused you people a bit. I also like to have the second option (I'm with Glen). But as far as the impl is concern, we can differ the way we should do it still leaving some flexibility

    [11:51] gdaniels: That's up to the individual implementation, dims - and you shouldn't need to "notify" it.... it just updates itself. :) [11:51] dims: ok, let's go for #2 then. [11:51] Ajith: ?

    [11:51] gdaniels: i.e. when the AxisConfiguration changes, that's all you need. Oh, it needs to recalc the handler chains, though. [11:51] dims: hoi dblevins [11:51] dims: exactly [11:51] Chinthaka: so, majority set for #2 ? [11:51] dblevins: dims: hi! [11:51] dims: with a notifier thing. [11:52] gdaniels: well, first let's make sure everyone understands everything. If there are more questions, let's discuss first. [11:52] gdaniels: then VOTE [11:52] gdaniels: is everyone OK on the differences/options? [11:52] Chinthaka: ok ... [11:52] dims: ok [11:52] Chinthaka: but I think we can't make a decision now .. [11:53] Chinthaka: better to put this in to the list [11:53] gdaniels: you think we should VOTE on the list, Chinthaka? [11:53] dims: we make a proposal to the list

    [11:53] dims: we all should agree first :) [11:53] gdaniels: ok [11:53] gdaniels: let's do both

    [11:53] Chinthaka: are we agreed ? :) [11:53] gdaniels: quick vote between us here and then we'll inform the list what we talked about

    [11:53] Chinthaka: me yes :) [11:53] dims: Let's do #2 (with minimal write?) [11:54] gdaniels: I think #2 as well [11:55] dims says it's getting real bad here outside my window [11:56] gdaniels: egads - I was just talking to Philippe about that... he was saying it's just ridiculous snow [11:57] dims: decision made? [11:57] gdaniels: ok yes

[15:12] saminda_ has joined

  • [15:12] Deepal: hI we are back [15:19] sanjiva_ has joined [15:19] sanjiva has disconnected: Read error: 110 (Connection timed out) [15:19] Chinthaka has disconnected: Read error: 110 (Connection timed out) [15:20] Ajith has disconnected: Read error: 113 (No route to host) [15:40] paulfremantle has joined [15:47] dblevins_ has disconnected: "This computer has gone to sleep" [15:47] dims: hi qll [15:49] jsisson has disconnected: Read error: 110 (Connection timed out)

    [15:52] dims: guys: FYI, see this - http://netzooid.com/blog/2005/12/09/5-cool-xfire-hacks/ [15:59] dims has disconnected [15:59] dims has joined [16:00] jsisson has joined [16:00] dims: PING [16:00] dims: oops. sorry about the caps lock [16:00] Chinthaka_: you wanna point us some cool methods one may like to have in Call ? [16:01] dims says me?

    [16:01] Chinthaka_: I saw another post in that blog, challenging Axis performance. I agree that we need to work on that too :(

    [16:01] Chinthaka_: http://netzooid.com/blog/2005/11/03/more-notes-on-soap-and-xfire-performance/ [16:01] dims: i'd like to revisit paul's OXM when we get a chance [16:01] dims: i haven't been able to spend time on perf yet [16:02] dims: it's on my TODO list. [16:02] dims: i'll be back in a couple of hours. [16:02] paulfremantle: ok [16:03] paulfremantle: see ya! [16:03] dims: sanjiva has my cell if you guys need something from me (which i doubt!) [16:33] Deepal: disussion on Call [16:33] Deepal: does call per an operation or service invocation [16:35] jsisson has disconnected: Read error: 104 (Connection reset by peer) [16:36] jsisson has joined [16:38] sanjiva_ is now known as sanjiva [16:39] Chinthaka_ is now known as Chinthaka [16:39] Ajith_ is now known as Ajith [17:13] dblevins has joined [17:15] dblevins has disconnected: Client Quit [17:16] dblevins has joined [17:22] jsisson is now known as HeWhoShallNotBeN [17:23] HeWhoShallNotBeN is now known as jsisson

    [17:25] gdaniels: Discussion re: engaging Modules and AxisService/AxisOperation/AxisMessage...

    [17:26] gdaniels: 1) The only handler chain that matters is the one in the MessageContext

    [17:26] gdaniels: 2) The flow when a new MessageContext enters the system is as follows... [17:26] gdaniels: * Copy the global "template" chain (pre-computed from the system defined phases) into the MC [17:27] gdaniels: * Run handlers until dispatch occurs [17:27] gdaniels: * When you dispatch to a particular point (service/op/msg), do something like this:

    [17:27] gdaniels: if (caching && cachedChain) { [17:27] gdaniels: copyChainToMessageContext(cachedChain, msgContext); [17:28] gdaniels: return; [17:28] gdaniels: } [17:28] gdaniels: engageMyModules(msgContext.handlerChain); [17:28] gdaniels: if (caching) { [17:29] gdaniels: copyChainToCache(msgContext.handlerChain); [17:29] gdaniels: } [17:29] gdaniels: ... [17:30] gdaniels: in other words, allow either computing the handlers by engaging modules, which will always be correct but may be expensive, or caching the computed chains and writing over the active chain with the cached ones (which doesn't account for dynamic policy). [17:41] csj has joined [17:41] csj: does axis c support attachments? [17:48] paulfremantle: good question.... i don't know if MTOM is in there yet [17:49] csj: paulfremantle, how about MIME [17:51] paulfremantle: well MTOM will use MIME... but MTOM is the approach to use now [17:52] paulfremantle: hold on [17:52] paulfremantle: let me just ask [17:52] paulfremantle: do you mean axisC++ 1.x or the new axis2/c stuff? [17:52] csj: either, just needs to support binary mime attachments [17:53] csj: I have a server side service in java that accepts mime encapsulated attachments, but I need a c / c++ client to talk with it.. [17:54] paulfremantle: ah ok [17:54] csj: I was using a java client for a while but for reasons beyond my control they need to be in c/c++ new [17:55] csj: s/new/now [17:55] paulfremantle: ok [17:55] paulfremantle: i understand [17:56] paulfremantle: hold on a min [17:56] csj: sure.. [17:59] paulfremantle: so... there is some limited support in AxisC++ 1.6 alpha [17:59] csj: still huh... server side only really huh? [18:01] paulfremantle: no both client and server [18:01] paulfremantle: but not sure how good it is [18:02] csj: are you a user or developer of axis? [18:03] paulfremantle: i worked on the axisc++ stuff but haven't been involved in it recently [18:03] paulfremantle: we are rebuilding it in pure C with a cleaner design [18:04] paulfremantle: but i can find out the status for you no problem [18:04] csj: that would be great, I dont mind helping to get attachments working but it seems as if the development team is a close group of people.. [18:05] paulfremantle: so the main contributors to AxisC++ are the IBM team [18:05] paulfremantle: they ship a packaged version of it:

    [18:05] paulfremantle: http://www-1.ibm.com/support/docview.wss?rs=171&uid=swg24009162&loc=en_US&cs=utf-8&lang=en [18:06] paulfremantle: whats the Java server? [18:06] csj: tomcat, my guess is your an IBMer :P [18:06] paulfremantle: no but i used to be [18:06] paulfremantle: is it Axis Java on Tomcat? [18:07] csj: yup..

    [18:08] paulfremantle: ok... so early next year we hope to support Axis2/C-->Axis2/Java with attachments [18:09] csj: ok.. [18:10] paulfremantle: if you give me an email i'll get back to you [18:10] csj: next year? [18:11] csj: well it is Dec :P [18:11] paulfremantle: no about the existing axisC++ support

    [18:12] paulfremantle: or send me an email to paul@wso2.com and i'll find out for you [18:12] paulfremantle: one of our guys was one of the key devs on AxisC++ 1.x and he will know [18:12] csj: ok, great.. [18:13] paulfremantle: no problem [18:14] csj: why are making the change to c rather then c++ [18:15] paulfremantle: ahha [18:15] paulfremantle: so we had a lot of people [18:15] paulfremantle: who [18:15] paulfremantle: had strange C++ environments [18:15] paulfremantle: or unnatural hatred of C++ etc [18:16] paulfremantle: and so since we had to do a lot of rework [18:16] paulfremantle: we figured C was more likely to be universally acceptable [18:16] csj: how is the c version going to support ssl? [18:16] paulfremantle: i.e. C++ users have no problems with C.. but not so clearly vice versa [18:18] paulfremantle: well we will probably put in an abstraction [18:18] paulfremantle: and then use openssl [18:20] paulfremantle: whats your opinion of the c/c++ issue [18:21] Chinthaka: can u guys please update


  • [18:21] Chinthaka: abt the issues we wanna do [18:22] csj: mm I dont think I really have an opinion.. :P [18:22] paulfremantle: ok [18:23] csj: I've looked at a great deal of the axis3 c++ transport, looks very good [18:23] paulfremantle: cool [18:25] csj: I just hope the c stuff is done in a way that I can drop in my own channels or whatever need be

    [18:26] paulfremantle: well the guy who wrote the c++ transport is leading the C work :-) [18:26] paulfremantle: well [18:27] paulfremantle: maybe not the 3 [18:27] paulfremantle: not sure [18:29] csj: when are they going to update the web page... still says they are using cvs thought they changed to svn [18:30] paulfremantle: no idea about that im afraid.... [18:30] paulfremantle: im off now - nice chatting to you [18:30] csj: thanks you too [18:31] paulfremantle has left

    [18:31] Deepal: http://wiki.apache.org/ws/Axis2/HackathonDec2005/chatlog

Axis2/HackathonDec2005/chatlog (last edited 2009-09-20 22:48:55 by localhost)