[JDEV] JECL library for Win32
Hi, Does anyone know if a win32 version of Jecl libary is available or under development? Don ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
Re: [JDEV] Perl Headline Delivery script Error?
You know you could talk directly with the developer who's code is in question instead
of asking a large selection of people that might not have even looked at Net::Jabber :)
Just a thought ;)
On Sat, 16 Mar 2002 02:46:43 +0530, "Gunjan Kakani" <[EMAIL PROTECTED]> wrote:
Hi,
I was experimenting with the Perl Headline Delivery script written by DJ Adams, from
"Fun with Jabber" articles.
But the script when executed gives error as following...
"Can't use string("3C924DDA") as a HASH ref while "strict refs" in use at
/usr/lib/perl5/./Jabber.pm line 588."
I checked with the changes made in the Net::Jabber module since this script was
written, but those changes could not solve this problem, since this problem occures in
sending the presence at $Connection->PresenceSend() method. If I uncomment this method
and produce the log, its working perfect, just that is is not sending messages to
subscribers, since they dont know the presence of the headline deliverer.
Anyone, has any ideas of this problem???
Thanks
-Cheers
-GC
___
jdev mailing list
[EMAIL PROTECTED]
http://mailman.jabber.org/listinfo/jdev
Re: [JDEV] Jabber.org vs. commercial Jabber structure
Michael F. March wrote: >>>- Multithreaded, supporting multiprocessor servers. >>> >>Although the jabberd-1.4.2 server has multithreading via it's use of pth, >>it's not "real" threads and cannot take advantage of multiprocessors (1.5 >>will no longer rely on pth and have this limitation). >> > >Has anyone tried to use this: > >http://oss.software.ibm.com/pthreads/ > >with Jabberd? One feature of this project is that it allows >pththread apps to use multiple processors on Linux. > The difference is in cooperative threading and preemptive threading - the 1.4.2 server was not written with preemptive threading in mind, and does not have locking to prevent resources from being trampled. NGPT is a combination of linux threads and pth threading (similar to how Solaris puts user-level threads on top of kernel-level threads). It is preemptive, so switching to it is not feasible without having resources protected against simultaneous modifications. -David Waite ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
[JDEV] Perl Headline Delivery script Error?
Hi,
I was experimenting with the Perl Headline
Delivery script written by DJ Adams, from "Fun with Jabber"
articles.
But the script when executed gives error as
following...
"Can't use string("3C924DDA") as a HASH ref
while "strict refs" in use at /usr/lib/perl5/./Jabber.pm line
588."
I checked with the changes made in the
Net::Jabber module since this script was written, but those changes could not
solve this problem, since this problem occures in sending the presence at
$Connection->PresenceSend() method. If I uncomment this method and produce
the log, its working perfect, just that is is not sending messages to
subscribers, since they dont know the presence of the headline
deliverer.
Anyone, has any ideas of this
problem???
Thanks
-Cheers
-GC
RE: [JDEV] Jabber.org vs. commercial Jabber structure
Hey, I'd advocate using something like ACE (http://www.cs.wustl.edu/~schmidt/ACE.html) as a basis for building new server type software. The reasons include: 1) includes an OS portability layer (which is ported to all major and many minor platforms). 2) Extensive research on high-performance servers, check out the publications 3) Very good abstract interfaces for common tools (tasks vs procs/threads) This has proven an excellent tool for at least 3 years, perhaps more. This could address alot of issues in the next server. --Ivan > -Original Message- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]On Behalf Of > Michael F. March > Sent: Friday, March 15, 2002 12:53 PM > To: [EMAIL PROTECTED] > Subject: Re: [JDEV] Jabber.org vs. commercial Jabber structure > > > > > - Multithreaded, supporting multiprocessor servers. > > > > Although the jabberd-1.4.2 server has multithreading via it's > use of pth, > > it's not "real" threads and cannot take advantage of > multiprocessors (1.5 > > will no longer rely on pth and have this limitation). > > Has anyone tried to use this: > > http://oss.software.ibm.com/pthreads/ > > with Jabberd? One feature of this project is that it allows > pththread apps to use multiple processors on Linux. > > ___ > jdev mailing list > [EMAIL PROTECTED] > http://mailman.jabber.org/listinfo/jdev > ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
[JDEV] Server-to-server connections down? Downgrading PTH will fix it!
Hi List! Just wanted to post a follow-up message. My jabberd servers were not working with server-to-server connections and I had to hunt through the archives to get the answer... so just wanted to say "Thanks" and post a "more obvious" subect-line-message to let everyone know that, yes, server-to-server IS broken when using libpth-1.4.1, but downgrading to 1.4.0 fixes the problem wonderfully! Thanks for having this great list, and great archives! Glen ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
RE: [JDEV] Jabber.org vs. commercial Jabber structure
> As a follow-up question, I don't think that we would ever have more > than say 10,000 concurrent Jabber connections at peak time. Is this a > load that version 1.4.2 could or has proven to be able to handle? I've never personally run a server that high, someone else on jdev might have more experience with those kinds of loads (it would likely take a good deal of local customizations and tweaking). So I can only say that for running the @jabber.org server it has done well with the <2000 loads. Jer ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
RE: [JDEV] Jabber.org vs. commercial Jabber structure
Jer, Thank you for the detailed reply. The pretty steep licensing price of the commercial Jabber prompted all our questions and great interest in the Jabber Software Foundation version. As a follow-up question, I don't think that we would ever have more than say 10,000 concurrent Jabber connections at peak time. Is this a load that version 1.4.2 could or has proven to be able to handle? Thanks, Mike ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
Re: [JDEV] Jabber.org vs. commercial Jabber structure
> > - Multithreaded, supporting multiprocessor servers. > > Although the jabberd-1.4.2 server has multithreading via it's use of pth, > it's not "real" threads and cannot take advantage of multiprocessors (1.5 > will no longer rely on pth and have this limitation). Has anyone tried to use this: http://oss.software.ibm.com/pthreads/ with Jabberd? One feature of this project is that it allows pththread apps to use multiple processors on Linux. ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
Re: [JDEV] Jabber.org vs. commercial Jabber structure
> I just browsed the FAQ and the many post of this mailing list to see if > my question was not already answered in the past. I didn't find it but > if somehow I missed it I would like to apologize to the group and, > well... invite you to throw the appropriate rocks at me. Heh, don't worry, jabber rocks are soft and fun, and fully parseable :) First, I just want to point out that "jabber.org" is the Jabber Software Foundation and the community of developers contributing to building the protocols and architecture for Jabber, not any particular server or project. In the past there was one primary open source server associated with jabber.org, but now that is just one of the numerous server implementations and can be found at it's new home of http://jabberd.jabberstudio.org/. > - Multithreaded, supporting multiprocessor servers. Although the jabberd-1.4.2 server has multithreading via it's use of pth, it's not "real" threads and cannot take advantage of multiprocessors (1.5 will no longer rely on pth and have this limitation). > - Added support for server farm architecture. There have been a few attempts at adding farming to jabberd, but none of them ever really caught on, possibly because the 1.4 generation of the codebase and pth wasn't the easiest to work with for such things (again, will likely be much easier to add to 1.5). That said, you can sort of farm a server by having many socket managers running on different servers, but they must all go back to a central session manager. > - Enhanced the number of simultaneous maintained sockets from 1,000 to > up to 25,000. As mentioned in the post just before this, there are a few other projects like dpsm and jpolld that can help manage more sockets, and the jabberd project has also moved to poll() for the future. > - Redesigned the internal database from a flat file to an optional > database By default there is only flat file support with jabberd-1.4.2, but there are numerous other contributions such as xdb_sql that move the internal data to a database. > - Enhanced reliability. > > What is the experience of this group regarding the number of concurrent > users that the Jabber.org server can serve? And what is its > reliability, Jabber Inc. claims that it is potentially poor needing > frequent restarts versus their version which supposedly is a lot more > reliable? The reliability part may have been true in the past, for the 1.2 release or earlier, but the 1.4 release and the recent 1.4.2 has been quite solid. I just checked and it's been running on jabber.org continuously over over a month now. All of this is only a small part of the picture and purely on a technical basis. I know that Jabber, Inc. has put significant time into designing and testing their server product, and has worked hard to make the setup, administration, commercial database integration, and scaleability work easily and with certian guarantees, they have done the hard part of the work that many businesses require. Meeting those same business needs with the various open source projects would require a good deal of time and expertise. Jer ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
[JDEV] Jabber.org vs. commercial Jabber structure
Hello, I just browsed the FAQ and the many post of this mailing list to see if my question was not already answered in the past. I didn't find it but if somehow I missed it I would like to apologize to the group and, well... invite you to throw the appropriate rocks at me. We would like to host a Jabber server as an added feature of our services and are trying to understand the load capacity of Jabber, the jabber.org popular version, and compare it with the commercial version published by Jabber Inc. It appears that Jabber Inc. went to great length to reorganize and optimize the Jabber server code in order to make it scalable to 100,000 and more concurrent users. In particular they made it - Multithreaded, supporting multiprocessor servers. - Added support for server farm architecture. - Enhanced the number of simultaneous maintained sockets from 1,000 to up to 25,000. - Redesigned the internal database from a flat file to an optional database - Enhanced reliability. What is the experience of this group regarding the number of concurrent users that the Jabber.org server can serve? And what is its reliability, Jabber Inc. claims that it is potentially poor needing frequent restarts versus their version which supposedly is a lot more reliable? Thank you for your passionate replies. Mike Kronenberg ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
RE: [JDEV] Max number of users....revisted!
The first limiting factor is the one you've tried to compensate for, which is that "select" is limited to FD_SETSIZE. Have you recompiled the jabberd source since the kernel tweaking? Also, another problem you might face is with the maximum number of open file descriptors any process can have at a given time (defaulted to 1024, varies by the OS). You can look into using ulimit to raise this threshold (ulimit -n 65536). That being said, instead of tweaking jabberd, you should take a look at the dpsm and jpolld daemons. They both offer ways to distribute the socket connection bottleneck. Basically, their only purpose is to manage client connections and funnel information to and from the main jabberd (via a single master connection to the server). The other advantage of these is that they both use "poll" to manage the I/O notifications instead of "select". Poll does not suffer from the FD_SETSIZE limitation, so the kernel tweaking isn't necessary (although, the ulimit tweak is still needed). Anyway, I hope that helps. Your results may vary. Jeremey Zela Northridge Systems, Inc. -Original Message- From: Riyaad Miller [mailto:[EMAIL PROTECTED]] Sent: Friday, March 15, 2002 9:46 AM To: Jabber Administration (E-mail); Jabber Developers (E-mail) Subject: [JDEV] Max number of usersrevisted! Hi all Small question ... What is the limiting factor on the number of simultaneous connections to jabberd? I'm currently using jabber 1.4.1 open source. It cries at 1000 (about) despite tweaking mgy kernel to get beyond 6 fd's. Are there any thread number issues involved? Could I run multiple jabberd processes on different ports to get a higher total number of connections to my server? Thanks Regards RM Give your child an unfair advantage with M-Web Learning. To join, call 08600 32 000 or go to http://join.mweb.co.za M-Web - JUST LIKE THAT ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
[JDEV] Max number of users....revisted!
Hi all Small question ... What is the limiting factor on the number of simultaneous connections to jabberd? I'm currently using jabber 1.4.1 open source. It cries at 1000 (about) despite tweaking mgy kernel to get beyond 6 fd's. Are there any thread number issues involved? Could I run multiple jabberd processes on different ports to get a higher total number of connections to my server? Thanks Regards RM Give your child an unfair advantage with M-Web Learning. To join, call 08600 32 000 or go to http://join.mweb.co.za M-Web JUST LIKE THAT ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
RE: [JDEV] AIM-Transport under cygwin
Just tried 0.9.24c and still got all the undefined references (in a more readable format albeit). Am I missing a library or something? It's not very clear... -Original Message- From: Brechtel, James Sent: Thursday, March 14, 2002 9:43 AM To: [EMAIL PROTECTED] Subject: [JDEV] AIM-Transport under cygwin Has anyone gotten this working? I've compiled jabberd 1.4.2 and pth 1.4.1, but when I try to compile the aim-transport (0.9.0) and it gets to building aimtrans.so I get undefined references errors like crazy. Anyone got this working before? ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
[JDEV] WAP Jabber Client
can any one tell me is there any WAP Jabber Client ? Satish Reddy ___ jdev mailing list [EMAIL PROTECTED] http://mailman.jabber.org/listinfo/jdev
