Hopbot log for 2008-06-23 - Helma IRC channel: #helma on irc.freenode.net

2008-06-23:

[18:50] <midnightmonster> re: the users list msg... anybody seen XML Blaster before? Besides the silly name, it sounds pretty cool.
[18:51] <zumbrunn> indeed, but I hadn't heard about it before either
[18:52] <midnightmonster> (and any clue why chanserv now thinks it's so friggin' important to +v people? what difference does that make?)
[18:52] <zumbrunn> it's not chanserv
[18:52] <zumbrunn> it's me
[18:53] <zumbrunn> is it annoying?
[18:53] <midnightmonster> it's a little extra noise and I don't understand what it does is all
[18:54] <zumbrunn> I just do it in order to easily be able to spot new users that joined
[18:54] <decke> *g* keep it up if it makes you happy...
[18:54] <midnightmonster> ok.
[18:55] <zumbrunn> but I can be without it
[18:55] <zumbrunn> ;-)
[18:55] <decke> midnightmonster: i have looked over xmlblaster from your mail -
[18:55] <decke> and it seems pretty interesting but it's not webservice like or?
[18:55] <midnightmonster> it doesn't matter to me. from my brief trying to figure out what it meant, it seemed to imply that if you didn't have +v you couldn't talk.
[18:56] <decke> currently i am researching in the area of webservices and WS-RM (WebServices Reliable Messaging) for an other project
[18:56] <zumbrunn> midnightmonster: that would be true if the channel is switched to be moderated
[18:56] <midnightmonster> it has a bunch of interfaces, including corba, xmlrpc, http, and email
[18:56] <zumbrunn> which we are not doing right now
[18:57] <midnightmonster> so it can be a webserice
[18:58] <decke> then WS-RM seems to be more suitable for my problem...
[18:59] <midnightmonster> ws-rm in google yields a lot of nothing
[18:59] <midnightmonster> in the sense of nothing that runs
[18:59] <decke> WS-RM is an WebService standard
[18:59] <midnightmonster> so what implements it?
[18:59] <decke> reliable messaging
[18:59] <zumbrunn> lol
[18:59] <midnightmonster> I think we have a language barrier
[19:00] <decke> no ws-rm is just an standard
[19:00] <midnightmonster> what software implements the standard? what can you run to use it
[19:00] <decke> ws-rm is currently implemented in .NET Framework 3.0 i think it's called something like Windows Communication Framework...
[19:01] <decke> on the server side i know of Sandesha2 (apache project)
[19:01] <decke> and there is Sandesha2/C which is pretty unstable but in C
[19:01] <decke> and a few more C/C++ libraries for possible clients
[19:02] <decke> sad but true - microsoft sucks - and they have missed the chance to put WS-RM into their Compact Framework
[19:02] <decke> but there are a few 3rd party .NET implementations for that
[19:03] <midnightmonster> possibly-relevant note from the FAQ about xmlblaster: "Note that there is no transaction support yet, but reliable store and forward (acknowledged message delivery) with crash recovery is available. For most applications this is quite enough. Transaction support is planned for a future release."
[19:04] <midnightmonster> also they have clients for nigh everything
[19:04] <midnightmonster> also, what is it with apache projects and incomprehensible names?
[19:05] <decke> yes but xmlblaster seems like the underdog in that case.. WS-RM has the big players behind it and is an officially supported standard
[19:05] <midnightmonster> "xml blaster"--corny but memorable, related to what it does, and easy to spell. "sandesha"--seriously?
[19:06] <decke> WebServices are a standard nightmare - one more project does not hurt anymore...
[19:06] <midnightmonster> (I'm just complaining about the names.)
[19:06] <decke> and Apache Axis might be more common to you if you know WS
[19:07] <decke> that's the WebService implementation from Apache - Sandesha2 is the WS-RM Plugin for Axis2
[19:08] <midnightmonster> I see all that now I've looked at the front page. I'm just saying the names stink. :-)
[19:08] <decke> *g* then post a bugreport for a project rename...
[19:09] <decke> but it's not that bad as "nefu" or "vagalume"
[19:09] <midnightmonster> anyway, "standard published by IBM, MS, BEA, and TIBCO" says to me: buckets of XML configuration, dozens of lines of code to do simple things, needs its own hardware to run.
[19:10] <decke> that's the good part about it... it's not that complicated
[19:11] <decke> let's say: _that_ complicated...
[19:11] <midnightmonster> (those names also stink in English, but it looks like "vagalume" may be more obvious in spanish or portugese)
[19:11] <decke> vagalume ~ butterfly
[19:12] <decke> an simple WS-RM sample is about 10 to 15 lines of Java code... so it's not that bad...
[19:14] <decke> there are a few examples... http://ws.apache.org/sandesha/sandesha2/userGuide.html
[19:15] <decke> but it gets a bit complicated if you need to have an app on your mobile device with Windows Embedded or something
[19:17] <decke> are you just interested in the conecpt of xmlbuster or is it for an knallgrau project?
[19:17] <midnightmonster> what's knallgrau?
[19:17] <midnightmonster> (so no, on that!)
[19:18] <decke> i thought you work at knallgrau... sorry my mistake ...
[19:19] <midnightmonster> I run a custom web shop / Micro-ISV.
[19:19] <decke> no i was not wrong... axel from knallgrau send that initial xmlbuster mail...
[19:20] <midnightmonster> yeah... that's not me.
[19:20] <midnightmonster> I'm joshua@papercrown.org who replied
[19:20] <decke> okay then it's just another mistake...
[19:21] <midnightmonster> (and crap, I just put my email in plain text in something that gets logged to a public page...)
[19:21] <midnightmonster> SPAM ME!
[19:21] <decke> it should not make a difference...
[19:22] <midnightmonster> ?I'm mostly interested in the concept. And very interested that XML Blaster offers a way for browser clients to subscribe via a persistent HTTP connection (COMET-ish)
[19:22] <midnightmonster> (Actually I have a very angry anti-spam proxy on my mail server that mostly handles such things for me)
[19:22] <decke> *g* you don't want to rape the http protocol - don't you?
[19:23] <midnightmonster> honest, your honor, it was completely consensual.
[19:24] <decke> i don't know an good concept that solves that http stateless problem...
[19:24] <midnightmonster> no, no... I'm not worried about statelessness
[19:24] <decke> but that was a protocol decision so if you have a problem with that - http not be the protocol you want
[19:24] <midnightmonster> it's just about getting those realtime updates
[19:25] <midnightmonster> if you have to talk to the browser, you don't get another protocol. gotta work with what you have,
[19:25] <decke> that's the same thing ... you want push from the server or?
[19:25] <midnightmonster> ?I have it on good authority that COMET is no violation of the HTTP protocol, just an unexpected application.
[19:26] <decke> i don't know COMET ... but there are only a few "conecepts" that give you http push...
[19:26] <decke> and they are all ugly as hell
[19:28] <decke> i have researched a lot on WebChat implementations that is the purest form of http push
[19:28] <midnightmonster> afaik, this is just, keep a connection open and send data only when you have it
[19:28] <decke> and you just can't get around it...
[19:29] <decke> that is only the common one yes...
[19:29] <decke> there are a few more...
[19:29] <midnightmonster> (I'm having a hard time conceiving anything besides polling and keeping a connection open.)
[19:30] <decke> you could have a long living connection
[19:30] <midnightmonster> isn't that what I said>?
[19:30] <decke> the client opens the connection - the server does not do anything with it... if it has data it sends it back... so you don't have to poll that often and have a better responsiveness
[19:31] <decke> but the client has to be threaded for that...
[19:31] <midnightmonster> which it's not.
[19:31] <midnightmonster> in the browser
[19:31] <decke> you can use JS for that
[19:32] <midnightmonster> you don't have threads in browser js, but you do have non-blocking IO
[19:32] <decke> but there are timeouts, caches and a lot of other problems
[19:33] <decke> the ajax ansynchronous requests would fit fine for that...
[19:33] <midnightmonster> anyway, so yes COMET feels like a hack, but if it's a hack you need, the idea of having a separate lightweight service that lets you "quarantine" the hack out of your main application is very appealing
[19:35] <decke> i don't know comet... is that a flash/java applet?
[19:36] <midnightmonster> no. it's just a catchy name (like AJAX) for occasionally sending data to the browser over a long-lived HTTP connection
[19:37] <decke> i got it... comet is exactly what i described above
[19:37] <midnightmonster> yes.
[19:37] <decke> long living connections... with a polling
[19:37] <midnightmonster> no polling afaik
[19:37] <midnightmonster> or just to reestablish if the connection breaks I guess
[19:37] <decke> "The browser makes an Ajax-style request to the server, which is kept open until the server has new data to send to the browser. After sending such an event, the server closes the connection, and the browser immediately opens a new one." wikipedia: Comet_(programming)
[19:38] <midnightmonster> ok.
[19:38] <midnightmonster> my idea of polling is that it's time based, not event based
[19:38] <midnightmonster> so I wouldn't call that polling
[19:38] <midnightmonster> but I see what you mean
[19:39] <midnightmonster> ?the biggest problem with implementing it is most web servers use up all their threads hanging on to the long-lived connections. more recent versions of jetty than ships with helma support comet, but most servers don't.
[19:40] <midnightmonster> it doesn't seem like it would mesh well with Helma's very traditional HTTP model.
[19:40] <midnightmonster> so I like the idea of offloading it to a lightweight other server
[19:41] <decke> that's the second time that that old jetty version in helma gets my attention...
[19:41] <decke> maybe it's time to raise it to hannes and ask for a reason....
[19:41] <midnightmonster> reason (I'm confident): it works already.
[19:41] <midnightmonster> and it would require some more work to update it with little perceived gain.
[19:42] <midnightmonster> (not that I wouldn't like to see it updateD)
[19:43] <midnightmonster> but my impression is it would be weird to use the comet support with the way helma is architected now
[19:43] <decke> the new jetty versions are much more powerful as far as i have seen it on the features and changes list
[19:43] <midnightmonster> even if we had an updated jetty
[19:43] <midnightmonster> (yeah--I would like to see an update)
[19:44] <decke> not just because they write it in big red letters in their documentation that AJP13 is bad and you better should not use it anymore...
[19:44] <midnightmonster> ?
[19:44] <midnightmonster> what should we be using?
[19:44] <decke> http
[19:45] <midnightmonster> what's wrong with ajp? I assumed it was a bit faster
[19:45] <decke> no that's no joke *g*
[19:45] <midnightmonster> (link?)
[19:46] <decke> http://docs.codehaus.org/display/JETTY/Configuring+AJP13+Using+mod_jk
[19:47] <decke> "Using the HTTP Connectors is greatly preferred, as Jetty performs significantly better with HTTP and the AJP protocol is poorly documented and there are many version irregularities."
[19:47] <decke> that's the chapter i was referring to...
[19:47] <midnightmonster> lol
[19:47] <midnightmonster> ok
[19:47] <midnightmonster> that's easy enough to switch
[19:47] <midnightmonster> I'm already using mod_proxy
[19:48] <decke> and it seems they made a lot of improvements to http only...
[19:48] <midnightmonster> just switch to _http instead of _ajp
[19:48] <decke> but with the older jetty it maybe just does not matter that much *g*
[19:48] <midnightmonster> right
[19:48] <midnightmonster> was just wondering how long ago it's true that http is better
[19:49] <midnightmonster> ???I'm also interested in xml blaster for its email distribution capabilities and as a possible foundation of a custom queue I'm going to need for another project
[19:50] <decke> just one last second to the helma jetty update...
[19:50] <decke> i have successfully ported helma to a jetty 7-pre
[19:51] <midnightmonster> cool.
[19:51] <decke> and delivered a few helma pages with it... locally and there were a few more bugs i have never fixed in that
[19:51] <midnightmonster> :-/
[19:51] <decke> but i would be possible
[19:52] <decke> i decided to put it behind because i don't know why it was never done... maybe it has some good technical reasons...

 

 

In the channel now:

Logs by date: