Menu
▾
▴
mod-pubsub-general — General list for mod_pubsub
You can subscribe to this list here.
| 2002 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
(7) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2003 |
Jan
|
Feb
|
Mar
|
Apr
(14) |
May
(12) |
Jun
|
Jul
|
Aug
|
Sep
|
Oct
(5) |
Nov
|
Dec
(2) |
| 2004 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
(4) |
Jul
(1) |
Aug
|
Sep
|
Oct
|
Nov
|
Dec
(1) |
| 2009 |
Jan
|
Feb
|
Mar
|
Apr
|
May
|
Jun
|
Jul
|
Aug
|
Sep
|
Oct
|
Nov
|
Dec
(1) |
|
From: Steven C. <st...@wr...> - 2004-12-17 11:38:42
|
Hi, I've gotten the repubsub python server up and running and have hacked at the test scripts trying to understand what's going on. It seems that the structure of an event is very much different from what is in the docs from the april 2004 tarball. It seems that events are javascript associative arrays and I can publish if I specify the method key as "notify" and I can specify a payload. But I want to append more data to the event like the rss scroller with s8_sitename etc. etc. However, when I add more keys to the event array, only the payload seems to get through. All the rest of the data disappears from the event array on the event handler side. Also, payload is reindexed as kn_payload. How do I structure events to pass data properly? Thanks Steven |
|
From: <mar...@ju...> - 2004-07-02 00:14:02
|
I am seeing what looks like a memory leak. I am using the latest version of pubsub.py from mod_pubsub (0.994) on Redhat (WS3) (and see the same problem on solaris). If I leave the server running for more than 3 days, ~50% of the memory is used by pubsub.py and the system becomes unresponsive. If I run the server and then run the perftool I will get an increase in memory usage as displayed by top. That number never goes down even after I stop perftool. I am trying to use mod_pubsub in a 24/7 application and having to reboot the box every day or two is not a very attractive option. Any suggestions? Mark ________________________________________________________________ The best thing to hit the Internet in years - Juno SpeedBand! Surf the Web up to FIVE TIMES FASTER! Only $14.95/ month - visit www.juno.com to sign up today! |
|
From: Mark B. <di...@ac...> - 2004-06-14 18:29:32
|
On Mon, Jun 14, 2004 at 10:43:13AM -0700, Adam Rifkin wrote: > I echo Joyce's sentiment, Mark. Go for it! Thanks guys, I will, I'm just not sure when. I'm still trying to decide between employment and another startup; if the latter, this will be a high priority. If the former, probably not so much, or at least not with the same urgency. Any pubsub.py gotchas I should be aware of, with respect to feature (in)completeness or (in)compatibility with the Perl version? Mark. |
|
From: <ad...@xe...> - 2004-06-14 17:44:47
|
I echo Joyce's sentiment, Mark. Go for it! Adam Joyce wrote: > Not that I know of. Most of the people who are interested in mps > right now are concerned with speed and scalability, not protocol. Go > for it! JP In response to Mark's writing: > Has anybody made any attempts at providing a cleaned up interface to > support a more RESTful use of HTTP? For example, using (many) kn_* > as HTTP headers, using the same request URI for GET and POST, a > "WATCH" or "MONITOR" extension method for non-GET routing, etc.. > > I'm trying to do some app-to-app stuff (no browser involved), and I'm > finding the protocol quite hairy. > > I'd be happy to work on this (on pubsub.py - Perl makes me ill), but > I thought I'd check first to see if anybody else was. |
|
From: Joyce P. <tru...@ya...> - 2004-06-13 16:54:34
|
Not that I know of. Most of the people who are interested in mps right now are concerned with speed and scalability, not protocol. Go for it! JP --- Mark Baker <di...@ac...> wrote: > Hi all, > > Has anybody made any attempts at providing a cleaned up interface to > support a more RESTful use of HTTP? For example, using (many) kn_* > as HTTP headers, using the same request URI for GET and POST, a > "WATCH" or "MONITOR" extension method for non-GET routing, etc.. > > I'm trying to do some app-to-app stuff (no browser involved), and I'm > finding the protocol quite hairy. > > I'd be happy to work on this (on pubsub.py - Perl makes me ill), but > I thought I'd check first to see if anybody else was. |
|
From: Mark B. <di...@ac...> - 2004-06-11 19:24:54
|
Hi all, Has anybody made any attempts at providing a cleaned up interface to support a more RESTful use of HTTP? For example, using (many) kn_* as HTTP headers, using the same request URI for GET and POST, a "WATCH" or "MONITOR" extension method for non-GET routing, etc.. I'm trying to do some app-to-app stuff (no browser involved), and I'm finding the protocol quite hairy. I'd be happy to work on this (on pubsub.py - Perl makes me ill), but I thought I'd check first to see if anybody else was. Mark. -- Mark Baker. Ottawa, Ontario, CANADA. http://www.markbaker.ca |
|
From: Joyce P. <tru...@ya...> - 2003-12-19 23:26:33
|
Hi Douglas, It would be super super cool if you ran the Python client on PocketPC! ASAIK, no one has done it yet, but as you say there's no intrinsic reason it wouldn't work. We do have a known-good PocketPC client written in C++, but your way sounds like way more fun. :-) JP --- Douglas Moore <dm...@ac...> wrote: > I was considering Python for a PocketPC 2002 implementation for a few > reasons: > Python is reported to be an efficient coding language > Python is portable to the PocketPC (again, so I'm told) > Python client for mod_pubsub is available... > > Does this sound like a good idea, has anyone done a PocketPC client > implementation or PocketPC and Python? > > Thanks in advance... |
|
From: Douglas M. <dm...@ac...> - 2003-12-19 22:09:24
|
I was considering Python for a PocketPC 2002 implementation for a few reasons: Python is reported to be an efficient coding language Python is portable to the PocketPC (again, so I'm told) Python client for mod_pubsub is available... Does this sound like a good idea, has anyone done a PocketPC client implementation or PocketPC and Python? Thanks in advance... Douglas Moore http://www.accius.com |
|
From: Joyce P. <tru...@ya...> - 2003-10-09 05:07:15
|
Danny! Where've you been? > So there's going to be a demonstration of installation? Why not have > some flow charts explaining how the server works? Awwww, you're mean. > Then you could say > that it is so easy to install even lazy government employees can get it > to work. After that I would do the whiz-bang parts. But I don't know any lazy government employees. I only know hard-working government employees who are maximizing the value of my hard-earned tax dollar through their contributions to world knowledge. JP ps How about some code? :-) |
|
From: Danny <dq...@nm...> - 2003-10-09 00:38:53
|
Hi Joyce, So there's going to be a demonstration of installation? Why not have some flow charts explaining how the server works? Then you could say that it is so easy to install even lazy government employees can get it to work. After that I would do the whiz-bang parts. Danny On Sunday, October 5, 2003, at 05:26 PM, Joyce Park wrote: > Hi Robert! So I imagine Apache httpd itself will be the framework for > this > particular road show, since that's our common ground. We'll walk them > through > setting up a mod-pubsub server -- or as I like to say, turning their > regular > Apache httpd into a pubsub server in three easy steps ;-) -- and then > show them > as many cool things as possible that they might do with that server, > and how > easy it is to do those cool things. -- Your advertisement here. http://hfun.org |
|
From: Robert O'B. <ro...@ih...> - 2003-10-06 10:12:51
|
Joyce Park wrote: > We'll walk them through setting up a mod-pubsub server... Which set-up; pubsub.cgi, modpubsub.pm and mod_perl, mod_proxy through to pubsub.py? > Most of the specific demos will be browser-based... Would some sort-of back-channel using trackback and/or a spin on the vote demo be possible as an interactive dimension to the talk. Instead of a saleforce.com dashboard an amazon/ebay market-place dashboard - wider market appeal than sfc. $.02 Robert. BTW: Does Google offer any sort of click-thru stats (via a service API) for advertisers? |
|
From: Joyce P. <tru...@ya...> - 2003-10-05 23:26:25
|
Hi Robert! So I imagine Apache httpd itself will be the framework for this particular road show, since that's our common ground. We'll walk them through setting up a mod-pubsub server -- or as I like to say, turning their regular Apache httpd into a pubsub server in three easy steps ;-) -- and then show them as many cool things as possible that they might do with that server, and how easy it is to do those cool things. Most of the specific demos will be browser-based, since presumably most Apache httpd developers will be interested in the browser; and we'll try to focus on specific emerging web needs where we can add value. We haven't decided on the exact lineup yet, but we'll try to touch on things of broad interest to webdevs -- like viewing Apache logs in realtime, non-email notifications, better RSS updates, and making your own front-end to multiple public web services. So to sum up, I think the narrative framework of this particular presentation will be "using a technology you already love, Apache httpd, to do cool new things that you're just starting to think about". Let us know if you can help out, JP --- Robert O'Brien <ro...@ih...> wrote: > Curious...what's the narrative framework for the ApacheCon talk/demo? |
|
From: Robert O'B. <ro...@ih...> - 2003-10-05 22:55:18
|
Curious...what's the narrative framework for the ApacheCon talk/demo? Robert. Joyce Park wrote: > at KN once which was supposed to model the activity of a > call-center -- and although it looked cool and showed off > the product well, we found that it entirely left people > cold because the "narrative framework" of the demo failed > to grab them. |
|
From: Lucas F. <lu...@de...> - 2003-05-22 21:44:16
|
Nevermind, got it. :) I had a javascript file open rathen than an html file, which apparently is read all at once by the browser... "Lucas Fletcher" <lu...@de...> wrote in message news:bajc41$diq$1...@ma...... > Hello, > > I'm attempting to write a simple .NET server as outlined in an earlier post > but am having problems with the browser executing the incremental javascript > code. The code is executed when the server connection closes, but not > incrementally as it is sent. (I am remembering to pad out each message to 4k > so as to flush the browser's buffer.) So I'm wondering if there might not be > something about the way .NET handles sockets that would cause this to be so. > (sniffing packets shows that the "push func" flag is set on the last packet > of each message for what it's worth...) > > Thank you, > > -- > Lucas Fletcher > lu...@de... > http://dealersinnotions.com > > > > > > ------------------------------------------------------- > This SF.net email is sponsored by: ObjectStore. > If flattening out C++ or Java code to make your application fit in a > relational database is painful, don't do it! Check out ObjectStore. > Now part of Progress Software. http://www.objectstore.net/sourceforge |
|
From: Lucas F. <lu...@de...> - 2003-05-22 20:36:39
|
Hello, I'm attempting to write a simple .NET server as outlined in an earlier post but am having problems with the browser executing the incremental javascript code. The code is executed when the server connection closes, but not incrementally as it is sent. (I am remembering to pad out each message to 4k so as to flush the browser's buffer.) So I'm wondering if there might not be something about the way .NET handles sockets that would cause this to be so. (sniffing packets shows that the "push func" flag is set on the last packet of each message for what it's worth...) Thank you, -- Lucas Fletcher lu...@de... http://dealersinnotions.com |
|
From: Joyce P. <tru...@ya...> - 2003-05-14 22:30:13
|
Hi Robert, I believe it is -- especially for high-volume uses. JP --- Robert Leftwich <ro...@le...> wrote: > Is the best way to ensure that a client has only one active > subscription > to a topic to specify the same kn_id when subscribing? > The problem I want to address is that, if I restart my (Java) > client w/o > restarting the pubsub (Python) server, I receive multiple event > notifications for the same event. If I add a specific kn_id when > subscribing then there is only one notification as desired. Is this > the > *best* practice for solving this problem? __________________________________ Do you Yahoo!? The New Yahoo! Search - Faster. Easier. Bingo. http://search.yahoo.com |
|
From: Robert L. <ro...@le...> - 2003-05-14 21:13:58
|
Is the best way to ensure that a client has only one active subscription to a topic to specify the same kn_id when subscribing? The problem I want to address is that, if I restart my (Java) client w/o restarting the pubsub (Python) server, I receive multiple event notifications for the same event. If I add a specific kn_id when subscribing then there is only one notification as desired. Is this the *best* practice for solving this problem? Robert |
|
From: Adam R. <Ad...@Kn...> - 2003-05-14 02:03:19
|
Lucas, This sounds awesome. I think your instinct that mod_pubsub would work with this, is right. Feel free to make the modifications -- and ask questions if you need any help. You might want to consider plugging this into the *Callback() functions in=20 mod_pubsub/kn_apps/kn_lib/pubsub_raw.js Not having to use frames would *definitely* be advantageous to some people. Thanks for bringing this up. Adam -----Original Message----- From: Lucas Fletcher [mailto:lu...@de...] Sent: Sunday, May 11, 2003 8:41 PM To: mod...@li... Subject: [Mod-pubsub-general] another way to initiate javascript streaming besides using frames I called a popup containing the script tag with the showModelessDialog method and it works great (in IE that is). It takes advantage of the = fact that the dialog is not displayed until the page is fully rendered which = of course it never is. No spinning world, totally transparent to the user. = I don't know if mod_pubsub would work with this, but with the right modifications I don't see why not... Not having to use frames might be = seen as an advantage to some people... -- Lucas Fletcher lu...@de... http://dealersinnotions.com ------------------------------------------------------- Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara The only event dedicated to issues related to Linux enterprise solutions www.enterpriselinuxforum.com _______________________________________________ Mod-pubsub-general mailing list Mod...@li... https://lists.sourceforge.net/lists/listinfo/mod-pubsub-general |
|
From: Lucas F. <lu...@de...> - 2003-05-12 03:52:00
|
I sent this to Dave, if anyone else is curious let me know... <dav...@on...> wrote in message news:B00...@vl...... > > -- > I'd be interested in reviewing your .NET server if you'd care to share it. > > I quickly whipped this out to try understand the reference to a continuous <script> file and it appeared to work fairly well. Of course IIS and .NET timed out which would negate the use of this combination of server tech unless this limitation could be circumvented. > > But check it out - I assume this is similiar to what KN is doing... > > http://www.bangeye.com/pubsub/ > > > Dave Cline > dav...@on... > www.bangeye.com/ > 801-636-5603 > > > > > -----Original Message----- > From: "Adam Rifkin" <Ad...@Kn...> > Sent: Fri, 9 May 2003 19:32:09 -0700 > To: "Lucas Fletcher" <lu...@de...>;<mod...@li...> > Cc: <mod...@li...> > Subject: [Mod-pubsub-developer] RE: [Mod-pubsub-general] javascript trick redux > > Hi Lucas, > > Browsers buffer data read off the wire. > > To overcome this, try sending 4k or so of spaces after the </script> tag. > > Also, we use hidden frames, not iframes. Have a look at the kn_debug > part of the "Command-line Arguments" section in > > http://www.mod-pubsub.org/kn_docs/javascript_reference.html > > or use ?kn_debug in the URL like so: > > http://www.mod-pubsub.org/kn_apps/presence1/?kn_debug > > and this will reveal those hidden frames. > > Hope that helps, > Adam > > > > -----Original Message----- > From: Lucas Fletcher [mailto:lu...@de...] > Sent: Thursday, May 08, 2003 10:27 PM > To: mod...@li... > Subject: [Mod-pubsub-general] javascript trick redux > > Hello, > > You thought you scared me away didn't you, with all your craziness and such? > Well it's not that easy I'm afraid. > > I've spent the last day or so writing a server in .NET framework that keeps > open a javascript file like you guys are doing. My friend who is hosting > 10's of thousands of flash-based clients (yeah, that's a custom server all > righty) says that the way to do it on Windows (he is using Linux) is with > I/O completion ports. His tests scaled up to at least a few thousands users > on one box. And I guess that the asynchronous methods in the > System.Net.Sockets namespace use I/O completion ports under the hood. This > uses the system thread pool for connection pooling. So I managed to write > the server with only a couple hundred lines of code. It uses a regular IIS > server for the upstream and then .NET remoting to communicate with the > custom server and then back down again via the javascript request. Which > brings me to a problem: the browser blocks execution on the javascript file > so the page doesn't load. Is there a special response header I need to > include? I can put the <script></script> tag after the closing <body> but I get some > weird errors elsewhere on the page, and also the IE world icon keeps > spinning. I noticed in the KnowNow demonstrations that this didn't happen. > Were there iframes being used? I covet its coolness. > > -- > Lucas Fletcher > lu...@de... > http://dealersinnotions.com > > > ------------------------------------------------------- > Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara > The only event dedicated to issues related to Linux enterprise solutions > www.enterpriselinuxforum.com > > _______________________________________________ > Mod-pubsub-general mailing list > Mod...@li... > https://lists.sourceforge.net/lists/listinfo/mod-pubsub-general > > > ------------------------------------------------------- > Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara > The only event dedicated to issues related to Linux enterprise solutions > www.enterpriselinuxforum.com > > _______________________________________________ > Mod-pubsub-developer mailing list > Mod...@li... > https://lists.sourceforge.net/lists/listinfo/mod-pubsub-developer > > > > > ------------------------------------------------------- > Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara > The only event dedicated to issues related to Linux enterprise solutions > www.enterpriselinuxforum.com |
|
From: Lucas F. <lu...@de...> - 2003-05-12 03:44:07
|
I called a popup containing the script tag with the showModelessDialog method and it works great (in IE that is). It takes advantage of the fact that the dialog is not displayed until the page is fully rendered which of course it never is. No spinning world, totally transparent to the user. I don't know if mod_pubsub would work with this, but with the right modifications I don't see why not... Not having to use frames might be seen as an advantage to some people... -- Lucas Fletcher lu...@de... http://dealersinnotions.com |
|
From: Adam R. <Ad...@Kn...> - 2003-05-11 00:23:22
|
Hi Dave, I will let Lucas answer as to whether he cares to share his .NET server. The problem with .NET/IIS closing connections is similar to the problem = Java application servers cause, too, in the name of scalability. = They're built to close connections as quickly as they can so they can = vend new incoming requests with greater quality of service. What you're doing at http://www.bangeye.com/pubsub/ is certainly similar = to what we're doing. Nicely done! Ka-Ping Yee demonstrated this technique using animated gifs almost four = years ago: http://zesty.ca/chat.html but when I go to that site today, it doesn't quite work. Other similar projects are: Pushlets - http://sourceforge.net/projects/pushlets/ http://www.pushlets.com/ OpenThought - http://sourceforge.net/projects/openthought/ http://www.openthought.net/ I'm Cc'ing Just and Eric on this note just to give them a heads-up and = thank them for teaching the world about HTTP streaming techniques. Adam -----Original Message----- From: dav...@on... [mailto:dav...@on...] Sent: Sat 5/10/2003 1:04 AM To: mod...@li... Cc:=09 Subject: [Mod-pubsub-developer] RE: [Mod-pubsub-general] javascript = trick redux --=20 I'd be interested in reviewing your .NET server if you'd care to share = it. I quickly whipped this out to try understand the reference to a = continuous <script> file and it appeared to work fairly well. Of course = IIS and .NET timed out which would negate the use of this combination of = server tech unless this limitation could be circumvented. But check it out - I assume this is similiar to what KN is doing... http://www.bangeye.com/pubsub/ Dave Cline dav...@on... www.bangeye.com/ 801-636-5603 -----Original Message----- From: "Adam Rifkin" <Ad...@Kn...> Sent: Fri, 9 May 2003 19:32:09 -0700 To: "Lucas Fletcher" = <lu...@de...>;<mod...@li...> Cc: <mod...@li...> Subject: [Mod-pubsub-developer] RE: [Mod-pubsub-general] javascript = trick redux Hi Lucas, Browsers buffer data read off the wire. To overcome this, try sending 4k or so of spaces after the </script> = tag. Also, we use hidden frames, not iframes. Have a look at the kn_debug part of the "Command-line Arguments" section in http://www.mod-pubsub.org/kn_docs/javascript_reference.html or use ?kn_debug in the URL like so: http://www.mod-pubsub.org/kn_apps/presence1/?kn_debug and this will reveal those hidden frames. Hope that helps, Adam -----Original Message----- From: Lucas Fletcher [mailto:lu...@de...] Sent: Thursday, May 08, 2003 10:27 PM To: mod...@li... Subject: [Mod-pubsub-general] javascript trick redux Hello, You thought you scared me away didn't you, with all your craziness and = such? Well it's not that easy I'm afraid. I've spent the last day or so writing a server in .NET framework that = keeps open a javascript file like you guys are doing. My friend who is hosting 10's of thousands of flash-based clients (yeah, that's a custom server = all righty) says that the way to do it on Windows (he is using Linux) is = with I/O completion ports. His tests scaled up to at least a few thousands = users on one box. And I guess that the asynchronous methods in the System.Net.Sockets namespace use I/O completion ports under the hood. = This uses the system thread pool for connection pooling. So I managed to = write the server with only a couple hundred lines of code. It uses a regular = IIS server for the upstream and then .NET remoting to communicate with the custom server and then back down again via the javascript request. Which brings me to a problem: the browser blocks execution on the javascript = file so the page doesn't load. Is there a special response header I need to include? I can put the <script></script> tag after the closing <body> = but I get some weird errors elsewhere on the page, and also the IE world icon keeps spinning. I noticed in the KnowNow demonstrations that this didn't = happen. Were there iframes being used? I covet its coolness. -- Lucas Fletcher lu...@de... http://dealersinnotions.com _______________________________________________ Mod-pubsub http://mod-pubsub.sf.net/ |
|
From: <dav...@on...> - 2003-05-10 07:04:42
|
-- I'd be interested in reviewing your .NET server if you'd care to share it. I quickly whipped this out to try understand the reference to a continuous <script> file and it appeared to work fairly well. Of course IIS and .NET timed out which would negate the use of this combination of server tech unless this limitation could be circumvented. But check it out - I assume this is similiar to what KN is doing... http://www.bangeye.com/pubsub/ Dave Cline dav...@on... www.bangeye.com/ 801-636-5603 -----Original Message----- From: "Adam Rifkin" <Ad...@Kn...> Sent: Fri, 9 May 2003 19:32:09 -0700 To: "Lucas Fletcher" <lu...@de...>;<mod...@li...> Cc: <mod...@li...> Subject: [Mod-pubsub-developer] RE: [Mod-pubsub-general] javascript trick redux Hi Lucas, Browsers buffer data read off the wire. To overcome this, try sending 4k or so of spaces after the </script> tag. Also, we use hidden frames, not iframes. Have a look at the kn_debug part of the "Command-line Arguments" section in http://www.mod-pubsub.org/kn_docs/javascript_reference.html or use ?kn_debug in the URL like so: http://www.mod-pubsub.org/kn_apps/presence1/?kn_debug and this will reveal those hidden frames. Hope that helps, Adam -----Original Message----- From: Lucas Fletcher [mailto:lu...@de...] Sent: Thursday, May 08, 2003 10:27 PM To: mod...@li... Subject: [Mod-pubsub-general] javascript trick redux Hello, You thought you scared me away didn't you, with all your craziness and such? Well it's not that easy I'm afraid. I've spent the last day or so writing a server in .NET framework that keeps open a javascript file like you guys are doing. My friend who is hosting 10's of thousands of flash-based clients (yeah, that's a custom server all righty) says that the way to do it on Windows (he is using Linux) is with I/O completion ports. His tests scaled up to at least a few thousands users on one box. And I guess that the asynchronous methods in the System.Net.Sockets namespace use I/O completion ports under the hood. This uses the system thread pool for connection pooling. So I managed to write the server with only a couple hundred lines of code. It uses a regular IIS server for the upstream and then .NET remoting to communicate with the custom server and then back down again via the javascript request. Which brings me to a problem: the browser blocks execution on the javascript file so the page doesn't load. Is there a special response header I need to include? I can put the <script></script> tag after the closing <body> but I get some weird errors elsewhere on the page, and also the IE world icon keeps spinning. I noticed in the KnowNow demonstrations that this didn't happen. Were there iframes being used? I covet its coolness. -- Lucas Fletcher lu...@de... http://dealersinnotions.com ------------------------------------------------------- Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara The only event dedicated to issues related to Linux enterprise solutions www.enterpriselinuxforum.com _______________________________________________ Mod-pubsub-general mailing list Mod...@li... https://lists.sourceforge.net/lists/listinfo/mod-pubsub-general ------------------------------------------------------- Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara The only event dedicated to issues related to Linux enterprise solutions www.enterpriselinuxforum.com _______________________________________________ Mod-pubsub-developer mailing list Mod...@li... https://lists.sourceforge.net/lists/listinfo/mod-pubsub-developer |
|
From: Adam R. <Ad...@Kn...> - 2003-05-10 02:38:10
|
Hi Lucas, Browsers buffer data read off the wire. To overcome this, try sending 4k or so of spaces after the </script> = tag. Also, we use hidden frames, not iframes. Have a look at the kn_debug part of the "Command-line Arguments" section in http://www.mod-pubsub.org/kn_docs/javascript_reference.html or use ?kn_debug in the URL like so: http://www.mod-pubsub.org/kn_apps/presence1/?kn_debug and this will reveal those hidden frames. Hope that helps, Adam -----Original Message----- From: Lucas Fletcher [mailto:lu...@de...] Sent: Thursday, May 08, 2003 10:27 PM To: mod...@li... Subject: [Mod-pubsub-general] javascript trick redux Hello, You thought you scared me away didn't you, with all your craziness and = such? Well it's not that easy I'm afraid. I've spent the last day or so writing a server in .NET framework that = keeps open a javascript file like you guys are doing. My friend who is hosting 10's of thousands of flash-based clients (yeah, that's a custom server = all righty) says that the way to do it on Windows (he is using Linux) is = with I/O completion ports. His tests scaled up to at least a few thousands = users on one box. And I guess that the asynchronous methods in the System.Net.Sockets namespace use I/O completion ports under the hood. = This uses the system thread pool for connection pooling. So I managed to = write the server with only a couple hundred lines of code. It uses a regular = IIS server for the upstream and then .NET remoting to communicate with the custom server and then back down again via the javascript request. Which brings me to a problem: the browser blocks execution on the javascript = file so the page doesn't load. Is there a special response header I need to include? I can put the <script> tag after the closing <body> but I get = some weird errors elsewhere on the page, and also the IE world icon keeps spinning. I noticed in the KnowNow demonstrations that this didn't = happen. Were there iframes being used? I covet its coolness. -- Lucas Fletcher lu...@de... http://dealersinnotions.com ------------------------------------------------------- Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara The only event dedicated to issues related to Linux enterprise solutions www.enterpriselinuxforum.com _______________________________________________ Mod-pubsub-general mailing list Mod...@li... https://lists.sourceforge.net/lists/listinfo/mod-pubsub-general |
|
From: Lucas F. <lu...@de...> - 2003-05-09 17:02:31
|
Hello, You thought you scared me away didn't you, with all your craziness and such? Well it's not that easy I'm afraid. I've spent the last day or so writing a server in .NET framework that keeps open a javascript file like you guys are doing. My friend who is hosting 10's of thousands of flash-based clients (yeah, that's a custom server all righty) says that the way to do it on Windows (he is using Linux) is with I/O completion ports. His tests scaled up to at least a few thousands users on one box. And I guess that the asynchronous methods in the System.Net.Sockets namespace use I/O completion ports under the hood. This uses the system thread pool for connection pooling. So I managed to write the server with only a couple hundred lines of code. It uses a regular IIS server for the upstream and then .NET remoting to communicate with the custom server and then back down again via the javascript request. Which brings me to a problem: the browser blocks execution on the javascript file so the page doesn't load. Is there a special response header I need to include? I can put the <script> tag after the closing <body> but I get some weird errors elsewhere on the page, and also the IE world icon keeps spinning. I noticed in the KnowNow demonstrations that this didn't happen. Were there iframes being used? I covet its coolness. -- Lucas Fletcher lu...@de... http://dealersinnotions.com |
|
From: Robert L. <ro...@le...> - 2003-05-07 10:19:42
|
Just to let you know that this mailing list is now available through the bidirectional email - nntp service at gmane (http://www.gmane.org). So now you can take advantage of all the features of newsreaders, such as threading, scoring, etc to help manage your ever growing volume of email from mailing lists. Robert PS Note that the archives are not available (yet). |
1 message has been excluded from this view by a project administrator.
