Archive-Date: XXX, 3 Nov 1994 14:14:17 GMT Subject: Re: [HELP] Would you help me? Message-ID: <39ar7p$ee9@nntpd.lkg.dec.com> From: Steve Huston Date: 3 Nov 1994 14:14:17 GMT References: <3985gd$s42@news.kreonet.re.kr> > I am looking for a package program named Phil Karn's KA9Q. If somebody tell me > where I can get it, it will be greatly appreciated. One place you can get it is anonymous FTP to gatekeeper.dec.com, in pub/net/ka9q There is a subdirectory in there named docs. I don't know what all the pieces are, or what to do with them after you fetch them, but there they are ;-) Steve Huston ================================================================================ Archive-Date: Tue, 08 Nov 1994 01:36:46 PDT Subject: Master Server TCP Message-ID: From: pp001529@interramp.com Date: Tue, 08 Nov 94 01:36:46 PDT Content-Type: TEXT/PLAIN; charset=US-ASCII I am currently using DEC TCP/IP Services (UCX V3.1) under Alpha AXP OpenVMS V6.1. I want to set up a "Master TCP Server" and have one "slave server" that manages each connection. The master server does the socket() and accept() calls, and I want to spawn another process, pass the file descriptor to it and do write() and read() calls. All this under VMS. We also are needing about 35 connections. I really don't care if I use QIO or C RTL calls. THe whole system runs under DCL, and there are lots more processes using global memory. How does one do this under VMS? Under UNIX I just fork() a process. Do I need to set something up in UCX> first besides the port services (like /etc/services)? I have heard things about AST, $ASSIGN and IO$DEACCESS. If this is part of a possible solution, please send code, or explain again. I saw an interesting thread but it has been deleted on my local site. BTW, we have no ideas if we are staying with UCX but would like to get the program working and evaluate performance options later. UCX seems to be supported on other products, and BSD socket calls seem supported also. ================================================================================ Archive-Date: XXX, 8 Nov 1994 15:52:35 GMT Subject: Re: Master Server TCP Message-ID: <39o6s3$ibq@nntpd.lkg.dec.com> From: MyTH@ucx.lkg.dec.com (M. T. Hollinger) Date: 8 Nov 1994 15:52:35 GMT References: In article pp001529@interramp.com writes: >I am currently using DEC TCP/IP Services (UCX V3.1) under Alpha >AXP OpenVMS V6.1. I want to set up a "Master TCP Server" and have one >"slave server" that manages each connection. The master server >does the socket() and accept() calls, and I want to spawn another >process, pass the file descriptor to it and do write() and read() >calls. All this under VMS. We also are needing about 35 connections. > >I really don't care if I use QIO or C RTL calls. We're working to get the sockets passed automatically to child processes when you use vfork(), but that functionality won't be available until the next version of the DEC C RTL. For now, if you don't care, it's probably easier to use the QIO interface. As long as the child process has the SHARE privilege or the parent process sets the UCX$C_SHARE socket option, you should be able to do I/O from both processes. Since you ask, I'll mail you a sample pair of programs (parent and child) which demonstrate this capability. One restriction, which we hope to ease in a future version, is that when the parent deassigns its channel, operations the child has pending on that device are cancelled. For example, if the child is waiting for input (blocked on an IO$_READVBLK QIO) when the parent deassigns its channel to the device, the operation will complete with SS$_CANCEL. To avoid this problem, you can either: 1) Have a handshake process where the child notifies the parent it has successfully assigned a channel, then waits for the parent to deassign it before doing any I/O. 2) Have the parent just leave the channel open until the child process terminates (as detected by a termination mailbox, for example). 3) Have the child process ignore the SS$_CANCEL code and just retry the operation. Note that there are other design options too, including the use of DECthreads (or your own multithreading approach) rather than using multiple processes. Yet another possibility is to have the child processes communicate with the parent through mailboxes, and let the parent handle all network I/O; that's sometimes appropriate when the child is doing a compute-intensive operation with minimal I/O over the network. - MyTH ================================================================================ Archive-Date: XXX, 2 Nov 1994 13:51:09 GMT Subject: [HELP] Would you help me? Message-ID: <3985gd$s42@news.kreonet.re.kr> From: nasol2@mgt.kaist.ac.kr (nasol) Date: 2 Nov 1994 13:51:09 GMT I am looking for a package program named Phil Karn's KA9Q. If somebody tell me where I can get it, it will be greatly appreciated. Also, I want any informations or comments, and if possible, other source programs for TCP/IP router(with respect to RIP, OSPF, SNMP(especially, MIB#2) etc.), which can be implemented and runned under "DOS" environment. Thank you ! Good luck with you ! from : Kim, dohoon Korea Advanced Institute of Science and Technology e-mail : kimdh@telmal.kaist.ac.kr date : 1994. 11. 2. ================================================================================ Archive-Date: Fri, 11 Nov 1994 19:43:20 GMT Subject: Re: Master Server TCP Message-ID: From: rustomji@swirl.monsanto.com (Eric Rustomji) Date: Fri, 11 Nov 1994 19:43:20 GMT Sender: news@tin.monsanto.com (USENET News System) References: In article , pp001529@interramp.com wrote: > I am currently using DEC TCP/IP Services (UCX V3.1) under Alpha > AXP OpenVMS V6.1. I want to set up a "Master TCP Server" and have one > "slave server" that manages each connection. The master server > does the socket() and accept() calls, and I want to spawn another > process, pass the file descriptor to it and do write() and read() > calls. All this under VMS. We also are needing about 35 connections. > > I really don't care if I use QIO or C RTL calls. THe whole system > runs under DCL, and there are lots more processes using global memory. > > How does one do this under VMS? Under UNIX I just fork() a process. > > Do I need to set something up in UCX> first besides the port services (like > /etc/services)? > > I have heard things about AST, $ASSIGN and IO$DEACCESS. If this is part of > a possible solution, please send code, or explain again. I saw > an interesting thread but it has been deleted on my local site. > > BTW, we have no ideas if we are staying with UCX but would like to get > the program working and evaluate performance options later. UCX > seems to be supported on other products, and BSD socket calls seem > supported also. Look into the system call SYS$CREATE_PROCESS (sort of equivalent to fork ) or similar and your master server can pass the connected socket as sys$input,sys$output and sys$error as parameters for the system call. I once had working code for this but I can't find it at the moment 8-( An even easier thing to do is just write your slave process and define it as a service in UCX. Then all you have to do is read/write to standard input/output/error Hope this helps you. Good luck. -- Eric Rustomji NSC Technologies Email: rustomji@swirl.monsanto.com Voice: (708) 506-2246 ================================================================================ Archive-Date: Thu, 10 Nov 1994 01:35:38 GMT Subject: Putting 100 people online Message-ID: From: mreardon@netcom.com (Michael Reardon) Date: Thu, 10 Nov 1994 01:35:38 GMT Could anyone recommend the best way to put 100 people (with there own terminals) on the net at one time, in the same room, and only for one or two nights? _____________________________________________________________________ mreardon@netcom.com -- mreardon@netcom.com ================================================================================ Archive-Date: XXX, 16 Nov 1994 19:23:18 +0200 Subject: Misc problems w TCPware Message-ID: <1994Nov16.192318.94@elmrd6.ineab.ikea.se> From: anos@elmrd6.ineab.ikea.se (Anders Ostling) Date: 16 Nov 94 19:23:18 +0200 Hi there Two problems; 1. Our DEC 7000-610 with VMS 6.1 and TCPware 4.0 does not show the correct IP client address in the SHOW USER/INTER/FULL and SHOW PROCESS displays. Instead of the correct A.B.C.D they are listed as A.B.C.C. Is this a known feature? 2. I have the DEC 7000 as primary nameserver and a Vax 3100-90 as secondary server. After modifying the NAMED.HOSTS (and updating the "serial"), I expect the changes to propagate. Not so. Sometimes, but not always. I should mention that the 3100 still runs V 3 of TCPware. I could include my NAMED.xxx files if that made it easier (8-)) to find the problem. BR Anders -- +-------------------------------------------------------------------------+ | Internet anos@ineab.ikea.se | | _ _ Voice +46-42-25 73 08, Fax 25 73 70, Attn: Anders Ostling | | \ \ \ IKEA Northern Europe AB, Sweden | | _/ _/ _/ | +-------------------------------------------------------------------------+ ================================================================================ Archive-Date: XXX, 24 Nov 1994 21:49:02 -0400 From: volz@process.com (Bernie Volz) Subject: Re: Misc problems w TCPware Message-ID: <1994Nov24.214902.839@process.com> Date: 24 Nov 94 21:49:02 -0400 References: <1994Nov16.192318.94@elmrd6.ineab.ikea.se> In article <1994Nov16.192318.94@elmrd6.ineab.ikea.se>, anos@elmrd6.ineab.ikea.se (Anders Ostling) writes: > Hi there Hi. > > Two problems; > > 1. Our DEC 7000-610 with VMS 6.1 and TCPware 4.0 does not show the correct > IP client address in the SHOW USER/INTER/FULL and SHOW PROCESS displays. > > Instead of the correct A.B.C.D they are listed as A.B.C.C. Is this a known > feature? > It turns out this is the result of a bug in the AXP MACRO compiler - it messes up the offset for an argument on the stack in certain cases and hence uses the wrong value in formatting the string. This is fixed in TCPware V4.1-3 by reworking the code to work around the bug. Contact support@process.com to request the V4.1-3 kit. > 2. I have the DEC 7000 as primary nameserver and a Vax 3100-90 as secondary > server. After modifying the NAMED.HOSTS (and updating the "serial"), I expect > the changes to propagate. Not so. Sometimes, but not always. I should mention > that the 3100 still runs V 3 of TCPware. I could include my NAMED.xxx files > if that made it easier (8-)) to find the problem. > I assume that by updating the "serial", you've increased the number (the value is larger)? More likely possibilities include: - Did you restart the primary name server? You need to do this after updating the database files. - What's the refresh and retry intervals on the SAO record? These control how often the secondary queries the primary to see if a zone transfer is needed. If these are long, decrease them (of course, a zone transfer won't occur until they do expire, so the values won't take effect until then). If these suggestions don't help, please contact support@process.com (and include the NAMED.xxx files). - Bernie Volz Process Software Corporation PS: Note that we're closed for the Thanksgiving holiday Thursday (11/24) and Friday (11/25).