Yellow Cards from 8/05/91 AIX 3.2 Migration Workshop MISCELLANEOUS (Brengle) ============= QUESTION: Who should we forward typos in Migration Guide to? ANSWER: Frank Hadden, AUSVMQ(HADDEN) QUESTION: What does gotchas mean? This is not included in my dictionary. ANSWER: Slang. Things that can "get you" or cause unsuspected problems. HARDWARE QUESTIONS (Brengle) ================== QUESTION: Multi-media adapter (audio capture/playback adapter): What compression ratios does it have on disk; i.e., how many MB per second? Pat Redmond ANSWER: At 80 pcm (adaptive differential PostCode Modulation) mono, the compression is 5,500 bytes/second. At 80 pcm stereo, it is 22,050 bytes/second. QUESTION: Single-chip RSC: size of instruction and data cache? Pat Redmond ANSWER: 8K byte cache, mixed instruction or data cache. QUESTION: New hardware (340, 350, 530H): Will they be orderable with 3.1 or 3.2, or ONLY 3.2? ANSWER: They can be ordered with 3.1. QUESTION: SCSI passthru terminator was NOT mentioned in hardware lecture yesterday. Is it in? Is it required for HANFS? ANSWER: Yes, SCSI passthru terminator is in 3.2. Yes, it is required for HANFS. The HANFS software will be a separate LPP. SOFTWARE QUESTIONS (Brengle) ================== QUESTION: Real-time support to conform to POSIX 1004? ANSWER: IEEE POSIX 1003.4, Real Time, is not a standard. It is a draft. It is IBM's intention ensure AIX meets relevant industry standards. When 1003.4 is a standard, you can expect that any real time support provided in AIX will adhere to it. QUESTION: How will HA-NFS be packaged? ANSWER: HA-NFS has been changed to HANFS (legal problems). It will be a separate LPP. QUESTION: How will XNS be packaged? ANSWER: In the bos QUESTION: How will TLI be packaged? ANSWER: TLI means Transport Layer Interface. It is a library shipped with the base operating system. QUESTION: What physical layers does XNS run over (tokenring, ethernet, X.25, ...)? ANSWER: ethernet and 802.3. NOT tokenring or X.25. QUESTION: What is the status on the Novell NetWare in AIX 3.2? Is it going to be announced as part of AIX 3.2 (net.obj) or will it be a new AIX 3.2 LPP? How will it be packaged? Which features will be included? Is all functionality included? ANSWER: Packaged as a separate LPP. It will operate like netware 3.1.1 in the PC environment. There is no netware console or NLM (netware loadable modules), but all other functionality is included. Console type support is provided through an application called sconsole, which is part of the LPP. QUESTION: In AIX 3.1, there is no facility to reduce the size of a filesystem. I was hoping this would be possible in AIX 3.2. Is this the case? ANSWER: No. QUESTION: Which LPPs are on the tape which the customers will receive? Do we need new releases of ALL LPPs to be installable on AIX 3.2? Peter Koepp ANSWER: All LPPs they order. The Pascal, Fortran, Fortran 2 and Cobol compilers are not AIX 3.2 products. However, the AIX 3.1 versions of these compilers will not install because they require a prerequisite of AIX 3.1. There will be an upgrade available which will allow these compilers to install on AIX 3.2. All other LPP's much be AIX 3.2 LPPs. BUG: INed craters sporadically by hanging up on hidden ("...xxx") files. Leonard James QUESTION: How does one have more than 16 virtual terminals on AIX 3.2? ANSWER: One doesn't. The chnumvt command only works for 1-16 hfts. Four is the current default number of terminals. QUESTION: Does AIX 3.2 PC Simulator support DOS Version 5? Fred Ho ANSWER: Not officially. OFFICIALLY, DOS Version 5.0 is NOT supported. However, development has tried running DOS Version 5.0 on pcsim and it appears to work. QUESTION: Is there a way to give a process unlimited CPU time? There are vendors who are developing applications, such as process control for process plants, that will require long running AIX processes. James Leonard ANSWER: AIX 3.2 now supports the use of "ulimit -t unlimited" as specified in the Kornshell specs. I understand that AIX 3.1 does not allow this. Under 3.1 you would do "ulimit -t 9999999999" and this would allow the process to run for about 40 years. QUESTION: If user exceeds his/her disk quota during a "quotaoff" period, what will happen after "quotaon"? Wodson Tseng ANSWER: The disk quota is only enforced when a user creates something. So if they are exceeding their quota when the quotaon command is entered, they will be unable to create a file or directory until they are below their quota. QUESTION: If I memory-pin a page retrieved by mmap, do I prevent it from being written through to disk? Lothar ANSWER: It is not possible to pin a page when it is being mmap'ped for user programs based on the current implementation. QUESTION: Most of the benchmarks that I ran, there is no difference in terms of double and single precision performance. Why? Exactly all of the cases (these are all floating point intensive applications). ANSWER: Try to recompile the benchmark with and without RNDSNGL option and compare the results. The benchmark should have large amount of FLOAT (instead of DOUBLE) calculation to demostrate the differences. NATIONAL LANGUAGE SUPPORT (Brengle) ========================= QUESTION: Are NLS programmers providing a set of files that can be deleted to SAVE SPACE if on En_US? ANSWER: No. The AIX 3.2 NLS packaging allows users to install only one language instead of all languages. QUESTION: How much has the NLS implementation impacted the size of the base OS? What can users remove to reduce the size (assuming they did not install all keyboards, message catalogs). ANSWER: Install only the language or languages that you need. We do not recommend that you delete any files. QUESTION: Which LPPs have/have NOT hard-coded messages into their programs for 'C' LANG? ANSWER: Products in AIX 3.1 which didn't have default messages to support the C locale will support the C locale in AIX 3.2. QUESTION: How much do the C messages differ from the En_US messages? ANSWER: The default messages in the operating system and the traditional unix commands is usually Unix-like: short, cryptic, assumes you know what it's talking about. The default messages in non-Unix products, such as SNA or HCON, are usually identical to the message catalog text, which is full sentences intended to ease translating the message into another language and to provide more information to the less-experienced user. INSTALLATION (Miller) ============ QUESTION: Is update a symbolic link to installp or is it completely separate? ANSWER: It is a link QUESTION: In 3.1.5, if you have two versions of the same LPP (e.g., Fortran 1.1 and 2.1) on the same stacked tape, it is not possible to install Version 1. If you try, it installs version 1 and then Version 2. Will this be fixed in AIX 3.2? Barry Graham, BASVM2(GRAHABM) ANSWER: If you say install all, you'll get all of the versions installed. Likewise, if you say to install xlf.all, you'll get all the versions. If you use the smit menu (or the comparable command), you can pick whichever version you want. QUESTION: What is match tape? ANSWER: It will only allow you to install LPP's installed on your 3.1 system. QUESTION: Why is the objrepos split up? ANSWER: Because of the new file tree structure. Those things that are unique to the machine are in /etc/objrepos. Things that are unique to the architecture are in /usr/lib/objrepos. Things that are available to everyone are in /usr/lib/share/objrepos. QUESTION: What specifically has changed in the ODM structures that will not allow 3.2 system to read 3.1 ODM configuration files? ANSWER: Nothing in the ODM structures. The ODM configuration methods have changed. QUESTION: Please provide specific information on why we can't use the 3.1 /etc/rc and /etc/inittab in a 3.2 system. You spend a lot of time telling us how to move it, but not WHY. What change changed?? ANSWER: The shell scripts to boot the system have changed because the flow of IPL has changed due to diskless. QUESTION: Inittab merged will contain devices that will no longer exist after installing 3.2. This will cause immediate errors on reboot. Tony Johnson ANSWER: We don't believe this to be true. QUESTION: With a preservation install, /(root), /usr and /tmp are overwritten. Since /var was, until recently, part of /(root), shouldn't /var be overwritten as well? Is it overwritten? Barry Graham ANSWER: /var is a new file system, not part of 3.1. So it is not overwritten, it is created newly it you're going from 3.1 to 3.2. COMMENT: Migration, merge files (i.e. inittab): before you do this, make backup copy of new inittab in case you mess up totally!!! QUESTION: The install process DOES recognize which disks contain the root volume group(s). How does this process tell the difference between root and other volume groups? Tony Johnson ANSWER: The install process interrogates all volume groups. It examines the filesystems in each volume group, looking for /, /usr, /tmp. If it finds them, it assumes it is a rootvg. QUESTION: Why do we use rmt0.5 and not rmt0.1? 5=low density, 1=high density. What is the difference? Tony Johnson ANSWER: On an 8mm, it doesn't make a different. On a 1/4" tape, there are different densities and it does make a difference. QUESTION: Can a customer do a preservation install if he has VGDA corruption? This preservation install probably does not recreate a VGDA, so what is the error message? ANSWER: VGDA=volume group descriptor area. It may or may not work. It depends on what the error is. QUESTION: The migration tools do not allow the end-user to migration any files that are not in the swconfig-info database without very convoluted means. You either have to manually back up files that are not in the database or you can extend the database (using odmadd), but those added entries will not be on the AIX 3.2 database, so that the old database must also be migrated. This is catch-22 situation. Shouldn't a tool be provided for the user to add their own list of files, and allow that list to be migrated easily. ANSWER: Tools are gone. QUESTION: lsswconfig -p option will list the new name of any file that does not exist in the previous release. Surely it would be best not to list files that do not exist in the previous release. Ray Hawkes ANSWER: Tools are gone. QUESTION: gtswconfig will not be able to build the software database from multiple media sources; e.g., if you have a stacked tape with the bos and OPPs and LPPs, but have OSI on a separate media, gtswconfig cannot pick up the OSI .mrg files for the database. Ray Hawkes ANSWER: Tools are gone. QUESTION: "SOFTWARE name" field on the "Restore Software Config" dialog uses a blank entry for ALL, whereas all other install menus/commands use the text "all". Shouldn't this field be changed so that it is consistent with other LPP install menus/commands? Ray Hawkes ANSWER: Tools are gone. QUESTION: The swconfig_info database does not have sufficient info to correctly transfer or merge files between versions. Example 1: if you have a combine file and you edit it to the correct context, the SMIT mvswconfig option will copy the files to the correct location, but the ownership/permissions on the file will be incorrect. Example 2: if you are going from 3.1.5 to 3.2, some files are replacable, such as /usr/lib/terminfo/ibm.ti, but from 3001 to 3.2, this may not be the case. The tools cannot cope with this situation. Ray Hawkes ANSWER: Tools are gone. QUESTION: Merge using diff: Why have both old and new of same name? Only have new one --> one you must have, i.e., brc!! Merge only those unique entries from AIX 3.1. ANSWER: Tools are gone. DISKLESS (Hadden) ======== QUESTION: Will the ODM hold diskless information (/usr/share/objrepos) or maybe in /export/root/client? Also, system dependent software info on client systems that belong in the ODM? Dan Edwards ANSWER: Each client will have their own copy of the ODM information. As each client in effect has access to "their own system code" they will have the ODM information for that code. In the case that the client will be sharing programs (LPP's) then the server for that program will have the ODM information. QUESTION: What is the limit on the number of diskless clients on one machine? ANSWER: Development is still testing at this point. They want to see how many they can get before it breaks. They will not speculate at this time. QUESTION: Will all models of the S/6000 be able to run diskless, as well as the 220? Barry Graham ANSWER: I am not sure I understand the question. If you mean, can all S/6000 serve diskless clients, the answer is all but the 220. If you mean, call all S/6000 be a client, the answer is only the 220 and the new desktops being announced at this time. QUESTION: If I increase swap space for a client, why is another file created instead of increasing the already-existing file? Lothar ANSWER: It is not supposed to create another file. It should increase the current file. Of course, you can create another file. I need to know exactly what commands they used to recreate this problem. QUESTION: Model 220 has built in ethernet - excellent! However, it uses thick net, so you still have to buy transceiver. Most of our customers use thin-net. Therefore, this built-in port is not as exciting as it sounds. Will there be a cheap transceiver available too? Barry Graham ANSWER: The 220 has built in ethernet, which is thick net. There are two converters. Feature code 4223 will convert from thick to thin net. Feature code 4224 will convert from thick to twisted pair. QUESTION: If you cannot have a model 320 as a diskless ystem, how can you have a 3.1 diskless system? ANSWER: You can not run AIX 3.1 on a diskless client, but an AIX 3.1 system can be a server for AIX 3.2 diskless client. Also OEM UNIX systems with the proper software can be a AIX 3.2 diskless server. QUESTION: Why can't you boot on a diskless from an ethernet card in the expansion slot? ANSWER: There is no code to support this in the diskless ros package. QUESTION: Diskless clients should not all have separate dump devices. This will be 24MG if you have 3 clients. ANSWER: This is an option that a system admistrator must make. If they have well behaved applications running there is no need for the excess dump space. But in the case that multiple systems sharing the same dump space have a failure at the same time you could end up with no useful output. QUESTION: Superclient definition on pg 11 in Frank's lecture - what do you mean by platform-specific code and same platform type? ANSWER: The superclinent is a term used when installing or maintaining diskless clients in a non-AIX 3.2 environment. If you need to have a user on a diskless workstation to be able to do installions, change permissions, and other types of work for AIX 3.2 that requires root or superuser privilges in an non-AIX environment then that person is defined as a superclient. (this is the same as having root or super user on the server) QUESTION: If this diskless goes '888', how ill the dump be transported across tcpip/nfs? Dan Edwards ANSWER: The network software will do the same as is done on a standalone system. If a dump is taken in a situation on the standalone system the diskless system will attempt the same. Should for some reason the network be down, then it can not respond to the dump request. QUESTION: Is tcpip/nfs functions pinned like sysdump calls, thus allowing "extended" functionality when the kernel panics (asserts)? Dan Edwards ANSWER: see above answer. QUESTION: Do we use the secure TFTP daemon for diskless? ANSWER: Yes, we are using secure tftp for boot image directory. FILE TREE (Hadden) ========= QUESTION: How strongly should we recommend that vendors change their code to reflect the 3.2 filesystem without using symbolic links? That is, if we push this strongly, will future versions/releases essentially preserve this filesystem structure? ANSWER: As other software vendors have already change their filetrees and the fact that were are working with OSF, AT&T, SUN, Berklery, and others in this matter we feel that this will be the way future systems will be shipped. Does this mean there will not be any future changes, NO. It is possible that in future releases that the symbloic links that we have to day will not be there. So this means that at some point if the user has application and/or programs that are relying on the old file tree and they at some time upgrade their systems they could have problems. In a lot of cases the applications may have already been on some other vendors platform and therefore they have made the necessary changes. GRAPHICS (Goldman) ======== QUESTION: Why do you have to recompile from Motif 1.0 to Motif 1.1? Aren't the libraries dynamic? QUESTION: How can we be binary compatible when OSF is not binary compatible on Motif? ANSWER: On AIX 3.1 system: Shared Motif 1.0 symbols are resolved to shr.o in libXm.a. On AIX 3.2 system: libXm.a contains shr4.o (Motif 1.1 symbols) with shr.o (Motif 1.0 symbols) following in order. If an existing AIX 3.1 Motif binary is executed, the symbols are resolved as before to shr.o in libXm.a. However if relinking occurs, the symbols will be resolved to shr4.o because it comes physically before shr.o in libXm.a. If you don't recompile your source you won't pick up changes in include files, data structures, and general API changes which will most probably break upon execution (relinking will work because symbols will get resolved). QUESTION: Display PostScript has how many fonts? If you could send me a list of these, I would appreciate it. Jean Veronneau, t/l 859-4358, VERONNEA at TORVM4 ANSWER: 22 fonts: AGaramond-Bold Courier-Bold Stencil AGaramond-BoldItalic Courier-BoldOblique Symbol AGaramond-Italic Courier-Oblique Times-Bold AGaramond-Regular Helvetica Times-BoldItalic AGaramond-Semibold Helvetica-Bold Times-Italic AGaramond-SemiboldItalic Helvetica-BoldOblique Times-Roman BrushScript Helvetica-Oblique Courier Hobo QUESTION: Can AIX graphics file translator/6000 generate a PostScript or HPGL type of file from a CGM file? ANSWER: No. QUESTION: X Server extensions: Why are the extensions not loaded dynamically? Load them whenever a program issues a call instead of making the decision when you start the server. ANSWER: The X Server base and extensions are integrated in certain structures in order to keep the server as lean and optimized as possible. If the extension was merged at application run-time this integration would not be as complete causing the X Server to be considerably larger than it already is. Also the X Server would take a substantial performance hit linking in the extension (impacting all active client apps most notibly your own). There are no plans at IBM or MIT to implement such a capability. GRAPHICS CORRECTIONS AND CLARIFICATIONS Pg 41 of Student Notes, next to last paragraph reads: "ICCCM is the selection ..." should read: "ICCCM defines the section ..." Compound Text: It was stated in the class that ICCC "owns" Compound Text Encoding. This is NOT true. While Compound Text is predominantly used in inter-client communications situations, it is it's own standard. It's really a much broader standard to address the issue of how to represent one piece of datum with its consitutiates encoded in several different codesets. ICCC says nothing directly about Compound Text but merely recommends a certain protocol that basically says: "When you transport data to/from other clients, make sure you tell the other client honestly what kind of data you are giving it and similarly, when you receive data, make sure you know what you are getting before you receive it." Without Compound Text, it would be virtually impossible to do this across vendor platforms (eg. HP doesn't know what ibm850 is). Concerning the 5 colormaps available on the POWER Gt4: As stated you have up to 5 colormaps concurrently available to you (default plus 4). Each one may be of any of the 3 X visual types made available by the server: PsuedoColor, TrueColor, and DirectColor. I stated that once a TrueColor visual was enabled (and a colormap used), subsequent TrueColor visuals would not use more colormaps because it is a read-only colormap. This is NOT correct. TrueColor visuals will continue to use Gt4 colormap resources as does the read-write type visuals. This is a redundant use of resources and will most likely be changed in the future. There is no graPHIGS version 2.2.2 on the 6000. It is refered to as the Personal graPHIGS within the AIXwindows 1.2 3D feature. As stated in the Migration Guide (pg 190) and the Student Notes (pg 111): - Version 2.2.2 (current announcement) is backwards compatible with V2.2.1 and V2.2.0. V2.2.2 requires AIX 3.2 and AIXwindows 1.2 base as well as the 3D feature. Changes to: - Personal graPHIGS within the AIXwindows 1.2 3D feature is backwards compatible with V2.2.1 and V2.20. INFOCRAFTER (Karner) =========== COMMENT: Try to persuade Interleaf to give IBMers free copies of its code to test InfoCrafter and promote it among customers. Pat Redmond, TORVM3(REDMOND) COMMENT: Filters are required to convert SGML, nroff/troff files (and other formats) into a form usable as input to InfoCrafter. nroff/troff is most important because many Unix applications will have man pages in this format that would be useful to import to InfoExplorer. Extra graphics filters would also be useful. Ray Hawkes, UK (HAWKESR@CROVM3) ANSWER: IDD's research showed that the most popular text formats for HARDWARE QUESTIONS (Brengle) ================== QUESTION: Does the 220 have ECC memory if it uses PS/2 memory SIMM's? Dennis Golden ANSWER: Yes, you can support ECC with the PS/2 memory SIMM's. SOFTWARE QUESTIONS (Brengle) ================== QUESTION: Why is it in 3.2, when you create a user, the login directory is pointing to /u. So, if I create a userid for myself (i.e., vmartin), when I login, my home directory is /u/vmartin. I would think that we should default to /home/vmartin with /u symbolically linked to /home. Vicki Martin ANSWER: Yes, it should probably be set to /home. It is too late to get a fix into AIX 3.2. However, there is a work around. You can modify /etc/security/mkuser.defaults to use /home instead of /u. COMMENT: Motif smit - "list" button excellent idea. Need better failure notification, perhaps color change. Also, reason for failure needs to be shown. ANSWER: Comment passed to development. It would require to DCR (Design Change Request) to change it at this point. The reason for a failure is displayed by the actual low-level command which is called by SMIT. If this is non-existent or inadequate, you can enter a defect on the specific low-level command, as SMIT just displays whatever is passed back by the low-level command. COMMENT: Put disk quotas in SMIT! Our customers that love smit (i.e., ease of administration) are the same people that have been demanding quotas (i.e., better control). ANSWER: Comment passed to development. It would have been in SMIT had there been time. QUESTION: Can smit give user the option to have "running person" or "word" icon? ANSWER: Not at this time. Development MIGHT add support for this in an update. QUESTION: motif smit: want to be able to continue using the tab key to look at choices ANSWER: We cannot use the tab key in Motif smit to look at choices because TAB means something to Motif. Part of the Motif look-and-feel is that Motif applications do not necessarily have to use the mouse. You should be able to see a black box around one of the buttons when you use Motif smit. That indicates where the keyboard focus is. If you were to press the enter key it would be just as if you had clicked your mouse in that button. Certain sets of buttons are placed in groups. The buttons in the smit menu are in one, the buttons on the bottom of the screen (Do,Cancel) are in another, and so on. These groups are called 'Tab groups' and if you press the tab key, you will move the keyboard focus from one group to another. You can try this in Motif smit. The arrow keys will move the keyboard focus around within a tab group and the tab key will move you to the next tab group. When the keyboard focus is on a Text field in a dialog (the little input window), you will have to press the tab key to move to the next field since the text widget has some special properties. This is how most Motif applications will behave and we chose to follow the Motif standard instead of the Curses smit method. QUESTION: motif smit: want to be able to look at the command before it is issued (commit fix menu didn't have this choice) ANSWER: You still can. Go to the 'Show' pulldown menu and select 'Command'. A popup window will appear with the command about to be executed. You can also press the 'F6' key if you don't want to go through the pulldown menu. QUESTION: motif smit: why was the old interface not supported also (e.g., arrow keys for movement in menus)? MSINCLR at WINVMJ. ANSWER: The arrow keys should still work in the menu. We tried to remain as consistent with Curses smit as possible but we chose to have Motif smit act as a Motif application. Motif specifies a set of behaviors for an application and we chose the Motif behavior over the Curses smit behavior if there was a conflict. Many of the function keys work the same in Curses and Motif smit however. You can find out more about using the keyboard by selecting 'On Keys' from the 'Help' menu in Motif smit. QUESTION: Will 3.2 be free upgrade for the existing customers? ANSWER: No. QUESTION: PRPQ P88635: 5080 emulation - RPQ's usually not supported. What support will come with this RPQ? ASKQ? Ingrid. ANSWER: Donna Van Fleet has decided that all PRPQ's originating in Austin will have full support. I called the Kingston developer for this particular PRPQ, and he tells me that this RPQ will also have full support from Kingston. I presume that means that if an ASKQ question is submitted, it will follow the usual path until someone (possibly a developer in Kingston or Austin) is able to answer the question. COMPILERS (Huang) ========= QUESTION: A vendor has requested for fixed point what FLTTRAP does for floating point. They want a compiler option that will insert code to check for fixed point exceptions such as overflow, and send signal if this occurs. Any plan to implement this, and what is the best way currently to detect fixed-point overflow? Thanks. Rich McCue, t/l 421-8532, RMCCUE@DALHQIC2 ANSWER: There are no plans for such an option now, and such support is partly a language issue. That is, we support FLTTRAP because the equivalent is available on other UNIX systems. However, fixed point exception detection is specifically not part of the C language, so we don't support that. I don't know if any other Fortran compilers support fixed point exceptions, or what the language definition says about such support. Languages such as ADA, on the other hand, demand that such support be available. Currently, it is very hard to detect fixed point exceptions in compiler generated code. Assembly code may be required. QUESTION: What is performance hit for using FLTTRAP? Thanks. Rich McCue, t/l 421-8532, RMCCUE@DALHQIC2 ANSWER: We have not done extensive performance analysis, but we expect 2X or 3X the execution time when FLTTRAP is used. This is very dependent on the type of code, however (ie, how floating point intensive, mix of float and fixed operations, etc.). QUESTION: I think there is a documentation error concerning SIGDANGER. Consider the consequences if the documentation were correct: handling SIGDANGER would allow a process to crash the system by exhauting page space. I haven't written the test case, but a customer claims that his process was killed the same, whether he handled SIGDANGER or not. In AIX 3.2, will it work as documented above, or will a single process crash the system? I think it is important to mention one or two things must happen when a process grabs page space without limit. Either the process must be killed or the system crashes (which should not be allowed). Thanks. Rich McCue, t/l 421-8532, RMCCUE@DALHQIC2 ANSWER: The SIGDANGER signal is generated when the system is low on page space, and means that running processes should attempt to reduce their paging space usage. Following the generation of SIGDANGER, if available page space gets even lower (meaning ANY process uses more), SIGKILL is generated. A process will not be SIGKILLed by the virtual memory manager if it has a signal handler to trap SIGDANGER. Note that this is not a good solution to the paging space problem -- the system will eventually crash if every process is equipped with a SIGDANGER handler. Your customer's claim does not fit what is described by the vmm implementation. It may be worth generating a test case to demonstrate the situation. When selecting the SIGKILL victims, AIX 3.1 picks the largest ones and removes them. AIX 3.2 has modified this behavior by choosing the youngest processes, one at a time, until the dangerous condition is over. Clearly, it would be desirable if just processes gaining a lot of page space were killed, but this will add another layer of decision making process to the vmm, and thus slow AIX runtime. As a summary, the major improvements are: 1) the vmm now picks the youngest instead of the largest (such as X server or some long-running job); 2) it now uses pistol instead of shot gun: only one process at a time is removed. NATIONAL LANGUAGE SUPPORT (Brengle) ========================= QUESTION: With introduction of NLS/ILS, will IBM change ordering systems (i.e., HONE, AES) to allow ordering of non-US systems/software in the US? This poses a problem for vendors/customers with both US and non-US locations. They cannot in the US order non-US hardware/software. ANSWER: One system is shipped worldwide now. A U.S. customer will have the Japanese locale available with the bos, just like a Japanese customer will. All locales are shipped with the bos. The user has the option of installing the ones they wish to use. Called Bob Carroll about keyboard. INSTALLATION (Miller) ============ QUESTION: There was not a need to run usetup. The environment was set up at the beginning of each of the other programs. ANSWER: usetup verifies that all of the software is available for using the upgrade utilities. Some of the commands that are required by the upgrade utilities are not standard aix commands--some may be included in the aix extensions. QUESTION: When we ran ". ./usetup" (step 6 in Installing the Upgrade Utilities) in a virtual terminal (created by "open ksh" on the hft), error message appeared: ... need to install ksh, awk, ..., ls, .... It is a bug? If not, it would be nice to issue a warning message at the beginning of the script. I believe running usetup in an aixterm result in the same error message. ANSWER: This is a bug. COMMENT: When aix is installing, it would be helpful for the list of devices to indicate which tape device is 1/4", 8mm, etc, so if you are unfamiliar with the machine, you don't waste time looking for the correct one. QUESTION: During the lab 3.2 upgrade process on Tuesday, a prompt appeared before installation. It stated the console could either be the 3151 or 5080. I chose the 3151 and after installation I changed the console to hft (5080). I then warm booted, but my console remained the 3151 with process messages on the 5080. When I shutdown completely via power off and rebotted, my console came up corrected as hft (5080). Jim Cargile, 8-4353 ANSWER: This is also a bug. DISKLESS (Hadden) ======== QUESTION: Assuming diskless is lost, can the installation and admin functions in 3.2 be preserved to provide better admin of "dataless" style implementations. Preserve PRODUCTIVITY tools. The "dwm" for dataless scenarios ... Keep /export and tools. Mark Johnson ANSWER: Latest rumor is that diskless isn't lost. But in the case the product is pulled I would think they would do like they have done in the past and that is remove all traces. This also one reason that it is still being worked by development. I am not sure just what tools you are talking about, but as most of them are shell scripts and do specific things for diskless, I am not sure they would do you much good. If you can provide specific commands I will see what their answer will be. (HADDEN at AUSVMQ) GRAPHICS (Goldman) ======== QUESTION: Will a terminfo be included for either xterm or aixterm? David Nation, DNATION@DAYVM2, t/l 269-6422 ANSWER: Yes, we ship terminfo entries for aixterm and xterm. They were shipped on AIX 3.1 also. GRAPHICS CORRECTIONS AND CLARIFICATIONS Pg 41 of Student Notes, next to last paragraph reads: "ICCCM is the selection ..." should read: "ICCCM defines the section ..." There is no graPHIGS version 2.2.2 on the 6000. It is refered to as the Personal graPHIGS within the AIXwindows 1.2 3D feature. As stated in the Migration Guide (pg 190) and the Student Notes (pg 111): - Version 2.2.2 (current announcement) is backwards compatible with V2.2.1 and V2.2.0. V2.2.2 requires AIX 3.2 and AIXwindows 1.2 base as well as the 3D feature. Changes to: - Personal graPHIGS within the AIXwindows 1.2 3D feature is backwards compatible with V2.2.1 and V2.20. INFOCRAFTER (Karner) =========== COMMENT: Many of my customers that would be interested in InfoCrafter use the FrameMaker package due to continuity of user interface (motif). The FrameMaker graphics output needs to be accepted by InfoCrafter (probably others as well). COMMENT: Please remove "Frame" from your InfoCrafter chart on page 14. ANSWER: And change it to Wordperfect? I want to give the impression of dedicated filters tied to a specific text format. This is not a pre-announce. It is "vox populi" that a Frame filter is in the works. Will check with Palo Alto, and if discouraged, will remove the reference. QUESTION: Why did the developers not allow integration of existing InfoExplorer documentation with that of user docs - i.e., allow a person to go through one path and one menu to get to multi-data. I know that question will keep coming up when I teach this. You wouldn't have to merge data - just allow pointers to make it look like user has to only go through one path. That should take care of the legal reasons. Thanks. Vicki Martin. ANSWER: As I memtioned in the lecture, there are 2 reasons, one legal the other technical. It is unclear whether the legal issue will ever be resolved to allow the merging (royalty scales are vastly different between IBM and non-IBM licenses). The technical issue is that dynamic binding is not supported. Will there be a good incentive in the future to do binds on the fly given the legal ramifications? Your guess is as good as mine. COMMENT: To avoid confusion, you should list docs of same type in your db1, db2, ... Example: src1.doc src2.doc src3.doc srn.doc ANSWER: This will give the wrong impression that dissimilar text formats cannot co-reside in the same directory. I dwelled long enough on the REQUIREMENT that the same text format must be used as input to the filter (as in the doclist file) but no such restriction exists for the directories and none should be implied or hinted either. SELECTIVE FIX (Williams) ============= QUESTION: Why happens to fix "info" after fix applied? Could it be retained as a database for problems? Could it be accessed via InfoExplorer (given InfoExplorer is a search engine)? Mark Johnson ANSWER: Fixinfo is kept available for each ptf that is applied. It can be listed with the lslpp -A command. QUESTION: Is it possible to provide a command or shell script that will prepare a summary of SW levels and/or fixes applied (a single command that provides all information that is required by IBM service)? Perhaps also SMIT screen .... can write to printer, file, diskette (potentially fax). Mark Johnson ANSWER: We will open a design record to request this function. I will also ask the change team to see if someting can be done for the interim. QUESTION: Are you going to include more information in RETAIN APAR and PTF descriptions? There has been a tendency to not include or have inconsistent information in RETAIN records. For example, level 2 centers may like to see what source modules are affected by a particular apar'ed problem. ANSWER: Will ask the change team to comply. QUESTION: Is it possible to supply fixinfo data into HONE? ANSWER: Will do. QUESTION: Feedback form shipped with the PTF's only rate the quality of packaging, media, etc. It should also rate the quality of the PTF itself. Dennis Golden ANSWER: Will forward this request to our distribution channels. It seems appropriate to ask the customer to rate the total package. REQUEST: We need documentation for selective fix support, whatever you will be sending to level 2. Please send to: Rich McCue, 41-03-60, Roanoke, Tx t/l 421-8532, (817) 961-8532, DALHQIC2(RMCCUE) Thanks. PROBLEMS: Problems with Selective Fix/smit problems - Could not get data to list off diskette using "all" option under "list all problems fixed by software" in smit - Interestingly, we could only get the software to commit using the "all" option. If we selected using a list, the software would not commit. - May want to point out that the order of applys and commits on output listing are reversed from AIX 3.1. May be confusing to some.