Sun-Spots-Request@RICE.EDU (William LeFebvre) (09/01/88)
SUN-SPOTS DIGEST Wednesday, 31 August 1988 Volume 6 : Issue 211 Today's Topics: Re: SLIP on SunOS4.0 Re: SunOS 4.0 Source Code licensing problem Re: botch in sunos4.0 ttytab. Re: Sunlink X.25 problems Re: Ethernet "balancing" when serving many clients Re: NFS write performance Re: RS-232 problem Re: Sun monitor/keyboard extensions -- summary Re: Clock losing 30 days Topologix making sunview program wait for a panel button click? coexisting Sun4.0 and Sun3.2 server? GNU Emacs SunView Interface Query Send contributions to: sun-spots@rice.edu Send subscription add/delete requests to: sun-spots-request@rice.edu Bitnet readers can subscribe directly with the CMS command: TELL LISTSERV AT RICE SUBSCRIBE SUNSPOTS My Full Name Recent backissues are available via anonymous FTP from "titan.rice.edu". For volume X, issue Y, "get sun-spots/vXnY". They are also accessible through the archive server: mail the request "send sun-spots vXnY" to "archive-server@rice.edu" or mail the word "help" to the same address for more information. ---------------------------------------------------------------------- Date: Mon, 29 Aug 88 21:39:59 edt From: oxtrap!rich@umix.cc.umich.edu (K. Richard Magill) Subject: Re: SLIP on SunOS4.0 Reference: v6n200 Brent Chapman writes: >An interesting rumor that I've been hearing... is that Sun's networking >group has SLIP on 4.0 up and running...and I would like nothing >more than for it to become an official, supported part of SunOS. No kidding. Only we didn't really want to buy anymore expensive sun SCSI drives when we had plenty of fast SMD sequent drives available and 4.0 does away with ND so... >Apparently, the Problem is that Sun already sells something similar >(called the Internetwork Router, or IR for short), and the marketroids >might not want to start "giving away" something that is almost as good as >IR, or slightly better, for many applications.... >I've never seen IR in action; does anyone use it?... Well... I've been trying to get INR to run over a 9600 leased line for almost a year now. I've been through sunos 3.4, 3.4.1, 3.4.2, 3.5, 3.5.1, INR level III, IV, and V as well as numerous patches from sun. IT STILL DOESN'T WORK!!! Don't get me wrong, we pass packets now and then but if we try to do anything reasonable like X, yp, emacs, make, sh, etc. we get flurries of network error messages. In the case of yp, both gateway 3/50's and the client and server machines become unusable (not to mention the link). But hey, at times I've clocked it at up to 750 char/sec gateway to gateway. Course, if we use dynamic routing and we fill the link, then routing packets don't make it through in time so the link gets marked down and well, you know. I figure either NO ONE ELSE in the world is using this product OR they don't show this problem (indicating either an error on our part or a unique configuration). Thus, sun doesn't really care whether it works for us. At this point I don't even want to hear from sun about it. It goes. We're pricing bridge boxes since we haven't seen SLIP for 4.0. BOTTOM LINE: I need something that works. INR DOESN'T WORK!!! On the other hand, its SUPPOSED to multiplex across several lines if you have them and I haven't seen slip do that. 'Course I haven't seen INR do much... Say... Anyone want to join me in an effort to commercially support SLIP? :-) -- rich. ------------------------------ Date: Tue, 30 Aug 88 16:51:45 PDT From: Craig Leres <leres@helios.ee.lbl.gov> Subject: Re: SunOS 4.0 Source Code licensing problem Mark Weiser writes: > SunOS 4.0 source apparently requires an AT&T System V rel 3.1 source > license. Notice that ".1". We have a Vr3 license (required for 3.5 Here's a relevant quote by Ozan Yigit from comp.unix.wizards: folks will be puzzling over "GNU: not enough core" messages on a 1024 meg sun5/3000. Doug Gwyn's rack will contain a 35-volume SXID (system-10 Interface Definition) that will be the latest AT&T-SUN sub-standard. [Sources will require a security clearence]. >From what I hear, it takes a real long time to get a security clearance, so I advise people to get started right away. Craig ------------------------------ Date: Tue, 30 Aug 88 15:38:52 EDT From: stpstn!aad@philabs.philips.com (Anthony A. Datri) Subject: Re: botch in sunos4.0 ttytab. We got a new 3/280 in a couple months ago, in a rack, from sun, ordered *without a bitmap*. The eeprom was set so that it booted from ie and *had a bitmap*, and the installation procedure didn't modify the ttytab to reflect the fact that didn't have the bitmap. Grrrrrrrrrrrrrrrrrrrrrrrrrrr. Anthony A. Datri,SysAdmin,StepstoneCorporation,stpstn!aad ------------------------------ Date: Tue, 30 Aug 88 17:53 BST From: Piete Brooks <pb%computer-lab.cambridge.ac.uk@nss.cs.ucl.ac.uk> Subject: Re: Sunlink X.25 problems Reference: v6n204 [ If you have no better replies ... ] > Sunlink X.25 release 5.0 and we are using the CPU serial port A for the > X.25 connection to PDN. I am testing the Sunlink by connecting two Sun > 3/50 back-to-back with a null modem cable I think it would be easier to get the link up connected to the PDN. As well as the sex of the connector and tx/rx (which the null-modem handles) you'll also have to get the clocks right (syncinit rxc/txc) & which end is the DTE and which the DCE (/etc/x25params hp_addr). [ I bolt a sex changer on the socket to make it a plug, and then make it a "DTE" in all senses. ] > Everytime the "x25config" program is run, called with our own tailored > x25params file, it stops after a pause with a message: > "X25_WR_SYSGEN: connection time out" I got this for a long time. From my first reading of the documentation, I took it that x25config simply read /etc/x25params and passed the info down into the kernel. However, it appears that it actually checks some levels, as you will get the time out message if there isn't a running X.25 driver at the other end (e.g. unplug the x.25 line). This sure makes starting back-to-back configurations hard ! ------------------------------ Date: Tue, 30 Aug 88 23:36:20 EDT From: hedrick@aramis.rutgers.edu (Charles Hedrick) Subject: Re: Ethernet "balancing" when serving many clients Mark wants to know whether it makes sense to use 2 Ethernets on a 3/260 to get better throughput in serving clients. Maybe, but I rather suspect you'll run out of disk throughput and CPU before Ethernet. Our tests show that with a single Supereagle and a Xylogics 451, a 3/180 uses at least 2/3 of its CPU. With a 3/280, things are a bit better but not that much. Basically we agree with Sun that the 3/280 gives you only about 1.5 the capacity of a 3/180 when used as a file server. This suggests that a 3/280 can probably handle 2 disk drives on separate controllers or on a single good disk controller (i.e. Ciprico or Interphase -- adding a second disk to a 451 doesn't give you any more performance, at least under SunOS 3.2. SunOS 4.0 may have fixed that). At that point I'd expect it to run out of CPU and disk bandwidth more or less at the same time. It's not clear to us that you will have saturated the Ethernet bandwidth by that point. You certainly won't have saturated the Ethernet itself. Maybe the controller. But I sort of doubt it. ------------------------------ Date: Tue, 30 Aug 88 11:19:13 PDT From: brent@sprite.Berkeley.EDU (Brent Welch) Subject: Re: NFS write performance > During the last several months I have noticed some rather serious > performance problems when writing files on my Sun workstations... > When I use the test program to write a file on a remote file system, I > find that the i/o rate on the server is many times the i/o rate reported > by the test program.... You guessed it. NFS servers do "write-through-to-disk" on EVERY client write operation. This includes updating the data block itself, plus the inode, plus any indirect blocks that are affected, and perhaps even the block allocation bitmap. That would give you the four-fold increase in disk activity on the server. This has big implications for paging to an NFS file, huh. The old swap partitions used ND, which is a device driver that implements a "Network Disk". ND was terrible for sharing (read only), but pretty efficient. I can't resist plugging Sprite at this point, our homegrown UNIX-like operating system we've been developing at Berkeley since 1984. We do delayed-write caching on diskless nodes and on the file server. The savings for things like temporary files, which live and die in the client cache, results in elapsed time savings of 30-40 percent on large makes. Of course, this also makes crash recovery more interesting, but not impossible at all. Our group does all its day-to-day computing on Sprite, the Sprite source code lives on Sprite, I am typing away on a Sprite client right now. We've finally hired a staff person (its been 4 of us grad students plus Ousterhout 'til now) and we're nearly satisfied with the implementation, so we hope to get a public domain release out "soon". We are, gag, doing a bunch of work now to make Sprite even more UNIX compatible in terms of library calls, etc. Oh, we run on Sun3's and Sun2's, and will port to Sun4's soon. Brent Welch UC Berkeley CS Division brent@ginger.Berkeley.EDU ------------------------------ Date: Tue, 30 Aug 88 15:10:29 EDT From: Paul Milazzo <milazzo@diamond.bbn.com> Subject: Re: RS-232 problem Charles Sandel writes: "I am having a problem getting my Sun-3/260 to read the output from a Heath 'Most Accurate Clock'. [...] the Sun just sees a bunch of nulls." I seem to recall that the Heath "Most Accurate Clock" provides a non-standard serial output because the clock has no negative voltage supply. Instead, they tie RS-232 signal ground to +5, and use 0 and +12 as signal voltages. Depending on how the SUN, Heath clock, and antenna leads are grounded, all sorts of horrible problems could result. The only real solution is to treat the Heath clock's output as a differential signal; perhaps you could just connect it through an RS-422 converter? Paul Milazzo <milazzo@bbn.com> BBN Laboratories Cambridge, MA ------------------------------ Date: Tue, 30 Aug 88 14:31:00 PDT From: paladin@sun.com (Joe Davis) Subject: Re: Sun monitor/keyboard extensions -- summary Reference: v6n205 >From: Daniel R. Ehrlich <ehrlich@blitz.cs.psu.edu> => Video pinout => => pin signal => => 1 Video+ => 2 Open => 3 HSync => 4 VSync => 5 Open <---this should say VCC ******* => 6 Video- => 7 Gnd => 8 Gnd => 9 Gnd => later, Joe ------------------------------ Date: Tue, 30 Aug 88 16:30:22 PDT From: Craig Leres <leres@helios.ee.lbl.gov> Subject: Re: Clock losing 30 days Matthew Turk writes: > Every time I reboot my 3/160, I get the message > > WARNING: clock lost 30 days -- CHECK AND RESET THE DATE! > > ...and sure enough it has. Anyone know how to fix this? Thanks.... > > [[ Checked your clock's battery lately? Maybe it's dead. --wnl ]] Wait a minute, isn't this a symptom of the yeap year bug in clock.c? Good God, the patch for that was posted enough times to enough newsgroups that I thought everyone knew it by heart. I think I got 4 hardcopies of it in the mail from Sun too. Also, if the battery was dead, I would expect the date to be more than 30 days off. Craig [[ Could be the leap-year creeping clock problem, although that usually made a clock run at a weird rate while the system was up. --wnl ]] ------------------------------ Date: 30 Aug 88 18:14:25 GMT From: crockett@amelia.nas.nasa.gov (Thomas W. Crockett) Subject: Topologix I originally posted this to comp.arch and comp.parallel, but got very limited feedback, so I thought I'd run it past the Sun folks: Has anyone out there had any experience with the transputer-based system built by a company called Topologix? There are four transputers to a board, and the boards can supposedly be plugged in directly to a Sun, limited primarily by the number of slots available. There is an operating system called LogixOS which runs natively on the transputers and on top of SunOS on the host. In fact, it is claimed that LogixOS can be used to run parallel applications across a network of Sun workstations, without even using the transputers. Is this a good product? I would be interested in experiences, both pro and con, as well any performance data which may be available. If you prefer not to post your comments publicly, you can e-mail them to tom@icase.arpa (alias tom@icase.edu). Thanks, Tom Crockett Institute for Computer Applications in Science and Engineering M.S. 132C, NASA Langley Research Center, Hampton, VA 23665 tom@icase.arpa (804) 865-4097 ------------------------------ Date: 30 Aug 88 18:24:44 GMT From: mkhaw@teknowledge-vaxc.arpa (Mike Khaw) Subject: making sunview program wait for a panel button click? How do I make a Sunview program pause until the user clicks on a panel button in one of the program's windows? If it matters, my Sunview program reads from/writes to a pipe connected to another process, and under certain conditions, I want the Sunview program to effectively block on reading the pipe until it sees the click on the panel button item. So far I've tried a wait loop. On the first try I called usleep() in the wait loop, testing for a condition that would have been made true by the PANEL_NOTIFY_PROC had the button click been received, but the program hung when it got into the loop. I then tried notify_set_itimer_func() in place of the usleep() call, with the same outcome. Mike Khaw -- internet: mkhaw@teknowledge.arpa uucp: {uunet|sun|ucbvax|decwrl|uw-beaver}!mkhaw%teknowledge.arpa hardcopy: Teknowledge Inc, 1850 Embarcadero Rd, POB 10119, Palo Alto, CA 94303 ------------------------------ Date: Tue, 30 Aug 88 17:15:45 PDT From: ametek!david@csvax.caltech.edu (David Lim) Subject: coexisting Sun4.0 and Sun3.2 server? We would like to bring up Sun4.0 slowly at our site also. Can anyone help me with the following questions? In all cases, assume Sun-3 machines. 1. Can programs compiled under Sun 3.2 run under Sun 4.0? I know this is a general question, but I'm trying to find out under what cases programs will not run under Sun4.0 2. Can programs compiled under Sun4.0 run under Sun 3.2? 3. We currently have our file server running Sun 3.2. In order to "slowly" bring up Sun 4.0, we plan to add another file server running under Sun 4.0. Has anyone tried this sort of arrangement? Thanks in advance! david@ametek.com ...!cit-vax!ametek!david ------------------------------ Date: 30 Aug 88 15:20:26 GMT From: Man/Machine Interface <mcvax!turing.ac.uk!mmi@uunet.uu.net> Subject: GNU Emacs SunView Interface Query We have a query about the SunView menu interface provided via 'Emacstool' and 'sun-fns.c' (Emacs 18.49 & SUN OS 3.4 EXPORT). Is it possible to modify 'sun-menu-evaluate' or, if necessary, 'sun-menu-internal' to allow an arbitrary user-specifiable function to be called every time the [highlighted] selection changes, i.e. every time a different menu item "goes inverse" when the mouse passes over it. Preferably this function would be given the string label of the newly highlighted item. Please note that this is rather different to doing something with an item once it has been selected, in this case we want to be able to perform our action[s] before any item is selected, while the user is still browsing over the options with the mouse. We would rather not rewrite the mouse-polling routines [... to compare the current mouse-position with the coordinates of the top-left hand corner of the menu that is being displayed every time the mouse moves and thus determine whether the item that is currently highlighted has changed ...] since this will obviously be too slow. Perhaps some kind soul with access to SunView sources could post an augmented version of 'menu_show' with this sort of functionality in mind. Ideally, the [highlighted] selection changing would be a visible event, allowing Lisp to do something (or nothing) when this event occurs. Obviously such a function would be the work of the kind soul concerned and could not possibly bear any relation to something that could be construed as being a breach of source licensing agreements and would never have been posted in the first place, had such been the case. ------------------------------ End of SUN-Spots Digest ***********************