IBM OS/2 1991 Technical Seminar

   What follows is a summary of the IBM OS/2 1991 Technical Seminar,

  for an audience I define as  "myself"  (what I would want to read

  if someone else attended in my place).

 

  This  seminar  is  to be repeated two more times (I think) in the

  USA,   so you can sign up for it yourself if you want.   Included

  with the seminar (to follow later in the summer) is a pre-release

  version of OS/2 2.00,   the IBM Programmer's Workbench,  SoftTerm

  (an asynch terminal program),    and  a  set of 32bit programming

  tools (including IBMs 32bit  C  compiler).    The price is approx

  $1595US.

 

  If your company will pick up the tab,  go for it.  If you're like

  me and have to fork out these dollars yourself (and  perhaps even

  more significantly: take time off of your current job to attend),

  I'm not sure I would recommend you attend.

 

  Notes

  -----

 

  Novel (3.2?) on AIX & OS/2 2.00 will ship sometime late 1992.

 

  ES/LS (Extended Services/Lan Server) will  run  on  other vendors

  OS/2 1.3/2.0.

 

  A new copy of OS/2 1.3 is US$150.  To upgrade  from  IBM  PC DOS,

  the cost is US$127 (and you KEEP your DOS license).

 

  The version of OS/2 2.00 they  had  at the seminar (for demos and

  for us to play with) was 6.140,  released 91/06/02.  It no longer

  has a Microsoft (c) copyright.

 

  OS/2  2.00  (hereafter  called  "V2")  will  "boot"  an arbitrary

  version of DOS  (such  as  DR  DOS  5.00 or CP/M) from a diskette

  image stored on your hard disk.  The VM (virtual machine) created

  for it thinks  it is booting from drive A:,  so you loose the use

  of physical A: for that session.

 

  If  you use the supplied DOS,  you get in excess of 647,000 bytes

  of free ram,   with XMS/EMS/MOUSE/NETWORK/CACHE/VDISK all drivers

  present.

 

  The current release TARGET for V2 is October.    This  is  only a

  target and is not an official release date.

 

  The ICON Editor comes with the base operating system.

 

  The FAT file system has been improved (coined "Super FAT"  by the

  presenter):

        - Lazy Writes.

        - Read Ahead Caching.

 

  V2  can  be  configured with "sticky keys",  so that a one handed

  operator can simulate multiple simultaneous key presses.

 

  The  session  manager  API is completely documented,  so that you

  can write your own shell program if you want.  You'll be told how

  to init PM and SPOOL.  The  example  given  for  this  was  for a

  customer wishing  to have a very controlled user workstation (one

  where users could not launch applications themselves).  I was not

  able to get the microphone to ask,  but this sounds like it might

  be a way to  boot  V2  without  PM  (I'm  not  sure  if  this  is

  desirable;    since  there is no such thing as a pure 32bit "full

  screen" program. ie: pure 32bit apps are always PM apps.).

 

  DosQProcStatus() will not be documented.   Mike Kogan (one of the

  better  speakers  at  the  seminar)  was quite amused that I  and

  others were able  to  find  and  use this undocumented API.  He's

  going to take the thought back with him as a user requirement. He

  said DosQProcStatus()  itself  will  never  be  documented,   but

  clearly we do want an API in there so that  we  can enumerate the

  running  processes.       He  told  me  to  use  DosQProcStatus()

  carefully, since it basically  stops  OS/2  while it collects the

  info it returns.   I brought the whole topic up when I discovered

  that the info  returned  from  it  is different from OS/2 1.3 ---

  another very good reason not to use it.

 

  WinAllocMem(), etc,  are  gone.   You're supposed to use the base

  OS memory functions,  such as DosAllocMem().

 

  There are some new PM Controls.  A CONTAINER control looks like a

  Windows 3.00 "Program group".   It supports icons,  bitmaps, text

  and date/time.

 

  A NOTEBOOK control looks like a spiral-binder with little plastic

  tabs  sticking out as bookmarks;  except that  the  tabs  can  be

  sticking out of any edge of the binder.

 

  The  SLIDER Control looks like something you'd  find  used  as  a

  "balance  control"  on your home stereo.   It is to  be  used  in

  place of a scroll bar which some people were using as a selection

  control,  rather than a scroll control.

 

  Palette Manager:  This new GPI feature allows a PM application to

  change the PM palette.  PM decides when to  change  the  palette,

  based on which application has the largest portion visible on the

  desktop,  and on which app is in the foreground.     Dithering is

  done for apps farther down in the z-order.  Simple  animation can

  be done by changing the palette.

 

  Popup Menus are menus that are not  tied  to  a  menu  bar.   For

  example,    a  user can go up to a menu bar,  and "drag" the Edit

  menu down to the middle of  their  document,  where it will stay,

  so that it is "within reach".

 

  The Font Driver Interface is documented.  This  is  the interface

  that  ATM  (Adobe  Type Manager; included with V2 and V1.3) uses.

  Therefore,  new font  engines  (such as TrueType) can be added to

  work with the font drivers supplied with V2.

 

  You can change the default font.

 

  There is a GpiQueryLocicalFont() API  (thank God!  You could only

  GpiSet...() it before).

 

  More VIO fonts.   128  character  wide  host session supported on

  VGA.

 

  Future Directions for PM:

       - 32bit PMWIN.DLL.

       - Distributed PM apps (32bit PM apps available  to  286 dos-

       windows workstations,  or non-intel platforms  such  as unix

       via X.11 protocol).

 

  If OS/2 and/or PM is ported to a new platform or technology (such

  as "NT"),  the 32bit API will remain the same.

 

  The  toolkit is going to come with  more  (and  better)  samples.

  (I'll believe  it  when  I  see it <g>).  The source code for the

  games included with V2 (for  example, an IBM developed version of

  Solitaire) will come with the toolkit.

 

  The 32bit C compiler was developed in house by IBM (in Toronto):

 

       - Full ANSI C compiler.

       - Re-entrant libraries.

       - Consistent across SAA platforms.

       - "lint like" options.

       - DLL support.

       - Optimizing.

       - Asynchronous and in-memory file i/o.

       - floating point libs.

       - Improved memory management.

 

       After I get  it  (Mid-July  time  frame) I'll tell you if it

       works or not <g>.

 

  V2 supports 16 bit apps using "LDT Tiling",  where  the 286-style

  pmode segments are mapped onto 386-style flat pages.  Thus the 16

  bit  apps are (ultimately) paging,  not  swapping.    The  entire

  16bit API is present in V2,  and is mapped onto the new 32bit API

  on the fly.

 

  Much of V2  is  still  written  in  16 bits (more on this later),

  even though the interface to it is the new 32 bit api.

 

  These  last two points lead to a discussion  of  "THUNKING".    A

  Thunk is a routine that translates a 16  bit  segment:offset pair

  into a 32bit flat offset,  or vice versa.   V2  relies heavily on

  thunking to support the old 16 bit apps,  and to interface 32 bit

  apps  with the portions of V2 which are  still  16  bits  (device

  drivers,  for example).

 

  Much thought was  put  into  this,   and thunking has very little

  overhead.  The 16bit selectors are assigned so that  a  simple  4

  instruction algorithm can be used to convert it to 32 bits,   and

  back again.  The  conversion  happens at ring 3 (ie: no privilege

  switch is req'd).

 

  Each  application  space  is 512MB.  This limit is not arbitrary;

  the LDT tiling and Thunking techniques limit it.

 

  Some Mixed model programs (programs that use both 32bit and 16bit

  APIS) may have to supply their own Thunks.  For  example,   a set

  of PM apps (one 16bit,  the other 32bit) that  want  to send each

  other      WM_USER     messages     will     have     to      use

  WinSetWindowThunkProc(hwnd,  pThunkProc)  to  establish  a  thunk

  procedure to covert any pointers  that are passed back and forth.

  (All system WM_* messages are autoThunked).

 

 

  I mentioned that V2 is still  made  up  of some 16bit code.  Here

  are some particulars,  as Mike Kogan told us:

 

       - The entire kernel is 32bits,  except for the file system.

       - There is  no  32bit DDI (device driver interface) defined,

       therefore all of the  device  drivers  use a 16bit interface

       (but may use 32bit logic internally).

       - Queues are still 16bit.

       - Window Manager (Win...() is still 16bit).

       - GPI/GRE is currently 16bit,  but will be 32bit by the time

       V2 ships.  (This is the  workhorse  of PM,  not the Win...()

       calls).

 

  GPI is currently coded in 286 asm,  and is being coded in 32bit C

  for portability (and performance).

 

  The XGA  driver  is  32bit,    exploits  bus mastering,  built in

  windowing support,  hardware managed pointer,  and can operate on

  system memory bitmaps.  (wow -- I can hardly wait  to  see one of

  those new 50MHZ 486 PS/2's running V2).

 

  Every  thread  now  has an exception handler  (something  like  a

  signal handler).  Thread 1 is a special case;  it handles process

  signals  and  the  Exit  List.     Killing  thread  1  kills  the

  application.

 

  Only two rings are used now,  ring 0 and ring 3.

 

  32bit apps can (optionally) let the operating  system dynamically

  grow  the  stack  of a thread,  thus very large  stacks  (in  the

  megabytes) are supported.

 

  Each VDM (virtual dos machine)  uses 4MB of linear address space.

  A  task switch from one to another is as simple as a register set

  switch, and a page-context switch (ie: it is very fast,  and does

  not require switching the CPU to real mode  as  is  done  in OS/2

  1.x).   The upper 3MB is defined as the "VDM workspace";  I don't

  specifically know what  it  is  used  for;  screen images and the

  like, I would guess.

 

  Each  VDM  has  access to 16MB of XMS memory, 32MB of EMS memory,

  and 512MB (508 actually) if it is a DPMI app (such as the Windows

  3.0 kernel that is included with V2).

 

  They assured me that the new semaphore API (for private/anonymous

  semaphores)  are  just  as fast (if  not  faster)  than  the  RAM

  SEMAPHORES that OS/2 1.x supplied.  The reason why  they  can  be

  (sometimes significantly) faster than the old RAM sems is because

  of a brain-damaged  DosClearSem()  in  OS/2  1.x.    I'll  try to

  explain;    if  several threads were blocked on a mutex semaphore

  under 1.x,  DosClearSem() would release ALL of the threads,   the

  scheduler  would  post  them all for execution,  figure out which

  thread had the highest priority,  then re-freeze all of the other

  threads.    OS/2  2.00 will only unblock one thread when a sem is

  cleared.     Apologies  to  Kent  Cedola  who  figured  out  that

  DosEnterCritSec(),  although not  as  elegant,  was,  much of the

  time,  faster than RAM sems.  Now we know why.

 

  16bit system semaphores are not compatible with any of the new 32

  bit  semaphores.    If a 32bit semaphore needs to use an app that

  uses  an  old  syssem,   it will have to be a mixed 32/16 bit app

  that uses the old API to access that particular sem.

 

  Named-shared  memory,    because  of  the 386 hardware,  is  much

  faster  than  OS/2  1.x,  since the kernel does not  need  to  be

  involved with memory transfers.

 

  Standard Dialogs are now  supplied  (thank  God!).  They actually

  look good too.  File Open,  File  Save,  and  Font  Selection are

  supplied.  They are customizable (no details given).

 

  OS/2 1.x had a separate thread (and process) for swapping;   thus

  swapping required a task switch.  V2 does not.

 

  There are four types of programs:

 

   link16-|      Pure 16 bit   |

   link32---|    Mixed 16 bit  |--- 16 bit runtime

            |    Mixed 32 bit   |

            |    Pure 32 bit    |--- 32 bit runtime

 

  If  you  take  a 16bit application,  and do nothing more than re-

  link  it,    it will become a "mixed 16 bit" EXE,  and will see a

  performance will increase.

 

  The new 32bit API is "c like" in  the  sense  that  arguments are

  placed on the  stack  from  right  to  left,  and the caller (not

  callee) cleans up the stack.   This  allows the api to be changed

  in the future  (by  adding  new  arguments  at  the  end) without

  breaking existing code (source or binary).

 

  MOSS  (I  think  this  stands  for   Multiple   Operating  System

  Selection) allows you to boot several operating systems from your

  hard disk.   It is an integral part of FDISK (FDISK,  by the way,

  now has a full screen version  as  well  as a PM version).  There

  are two ways  you  can use it:  1)  have more than one drive "C:"

  on your hard disk,  such that booting  one  of  them  "hides" the

  others.  This is  useful  for  something  that insists on booting

  from C:.  2) Boot from another partition,  such as D:.  OS/2 2.00

  will boot from D: (yiiiiiiippppppppeeeeeeee).  This  later method

  lets  you access the partitions (such as  C:)  that  may  contain

  other bootable operating systems.

 

  V2 install allows arbitrary source and destination drives.

  It will install over a LAN.

  It has a graphical install program.

  It uses PKZIP technology to speed up installation.

  You can jump to a command prompt at any time during Install.

 

  CSDs:  Graphical interface,   Lan  support,  Progress indicators,

  Backout procedure.

 

  Recommended Minimum Configuration:

 

                      1.2       1.3       2.0

                      -----------------------

       SE             3-5mb     2-4mb     3-5mb

       LAN Requestor  4-7mb     3-6mb     4-7mb

 

  Applets: Calender and Diary,  Address Book,  Mini-spreadsheet and

  Database,  Calculator,  Alarm Clock,  Business Graphics charting.

 

  The next beta will have "32bit protected DLLs".  Currently (under

  1.x),   a DLL that has SHARED data per instance can have parts of

  its  dataspace corrupted by a separate  processes.  (It  may  not

  happen often,  but it IS possible).   This  can  even  crash OS/2

  itself.   These new 32bit protected DLLs  will  be  able  to have

  shared data (and/or  be  of  the  "init  global"  flavor)  and be

  uncorruptable.

 

  A little more about MVDMs

  -------------------------

 

  Migrate data to/from via  PM  clipboard  (for  win and normal dos

  text apps).

 

  Named pipe support.

 

  Bimodal device drivers (real/protected)  are  no  longer required

  (This makes the job of writing an OS/2 DD a lot easier).

 

  All but 10k of the  640k is available for a DOS app.  This 10k is

  primarily used to support many of  DOS's  undocumented structures

  and features.  (note: The device driver chain  is  not supported;

  not many programs use this --  some  network programs do.  If you

  want this level of compatibility,  simply boot a real  version of

  dos in a VDM).

 

  Hardware  is  simulated.    The  following  VDDs  (Virtual Device

  Drivers) are supplied:

       - BIOS

       - PIC (interrupt controller)

       - EMS

       - XMS

       - DMA (allows copy protected programs with a  "key  disk" to

       work,  as well as programs like FASTBACK).

       - Virtual CMOS

       - NPX (80x87).

 

  Some  of  us  speculated that V2  would  contain  a  BCL  (Binary

  compatibility layer).  Apparently  this  was  coded,   but it ran

  like a dog.  Since IBM has a full cross-license for  Windows 3.00

  source,   they simply included it with V2.    They  have  removed

  large chunks of it  (dealing  with  "386  enhanced mode"),  since

  these features are now supplied by V2 itself.   They  also turned

  Windows 3 into a  DPMI  (Dos Protect Mode Interface) client,  and

  it is run  as any other DPMI client is under V2;  except there is

  a few hooks in there to support cut/paste for the clipboard.

 

  Win 3 runs in "standard mode" under V2.  You can  therefore start

  more than one winapp in a win  3  session,    and  they  will use

  "cooperative muktitasking" <giggle>,  and crash  each  other just

  like they do under real Windows  today.    Or,  you can start one

  winapp per Windows 3 MVDM,   and get pre-emptive multitasking and

  full memory protection.

 

  Ron Cadima did a side by side V2/Win3 demo for  us.   The Windows

  machine had twice as much memory installed (16mb vs  8mb  on  the

  OS/2 2.00 machine).    The  Win3  machine was configured to be as

  fast as possible (large  smartdrv,  himem,  enhanced  mode, etc).

  The V2 machine  was  using  the  new  SUPER FAT (not hpfs or lazy

  write),  so  as  not to make Win look too bad <g>.   Two programs

  were used for this demo;  a Win3 "disk write" program that writes

  data out to disk a few  hundred  times,   and a Win3 program that

  draws a MAP about a dozen times.   Both  programs  time their own

  execution.

 

                                Windows   OS/2

                                --------------

       DISK WRITE               28        10

       DRAW MAP                 4         8

       DRAW MAP/DISK WRITE      26        8   (time for map)

 

  (time in seconds).

 

  You'll notice that the win machine did one thing  faster  --  the

  drawing of the  map  while  no  other programs were running.  But

  look  at  the  difference when the same program was running while

  the disk write program was running!.

 

  REXX

  ----

 

  Not much new here,    except for performance.  When they coverted

  it to 32bit,  they saw a significant (60%?) performance increase.

 

  There will be an online REXX reference.

 

  IBM  really  wants  this to be the next "basica",  and spent more

  time talking about it than I wanted to hear. <g>.

 

  Programmer's Workbench

  ----------------------

 

  This is a PM program that is extensible.  It reminds me very much

  of Microsoft's PWB program  (which  is NOT a PM program),  except

  that it  does  not  have  a  built  in editor.   There was a demo

  given,  but  it  is  hard  to  judge such things without using it

  yourself.   I saw nothing revolutionary about it.

 

  Actually,  I found IBMs hyping of this tool  a  little  arrogant.

  They  describe  it  as  the  last  IDE   (integrated  development

  environment)  you'll  ever  use,    and  "of  course" all  future

  compiler vendors will  support  it.      Ten  minutes  later they

  touched on Borland C++,    which will ship with it's own Borland-

  standard  IDE  (but  will probably plug into IBMs workbench as an

  option).

 

  Except for the  possible  interface to IBMs workbench,  Borland's

  C++/TASM for  OS/2  has  nothing  to do with IBM.  It will not be

  licensed or marketed by IBM.

 

  Oddly enough,  Zortech  was  one of the guest participants at the

  seminar,  yet they  were  not mentioned in the slides during this

  presentation.  One  of  the attendees (obviously an angry Zortech

  C++ fan) pointed this  out  to the presenter during Q&A.  Zortech

  ships a C++  compiler  for  OS/2  today.  Borland's has a planned

  (but not announced) ship date of 2H92.   Go figure.

 

  The Borland slide  had  "PM Debugger, PM Profiler, OOP Assembler,

  Resource editor and compiler" listed as features.  Could be good!

 

  IBM, by the way, will be shipping a PM debugger with  their 32bit

  C compiler.      It  will have a built in PM queue debugger (like

  PMSPY).

 

  IPF

  ---

  The Information Presentation Facility will have author-controlled

  fonts and horizontal scroll bars.

 

  Multi-view  support  (for mixed text/graphics  and  motion  video

  support, etc.).

 

  Extended Services.

  ------------------

 

  A lot of time  was  spent  talking  about DBM (database manager).

  The most significant new feature is "roll forward".  All and all,

  DBM is still not as good as SQL Server in  my  opinion --- unless

  you are planning to interface to other SAA platforms.  (There are

  ways to do  this  with  SQL  Server  too,    though.  Not all SAA

  platforms,  but you never know).

 

  There are no plans for an SQL API; we're stuck with  the embedded

  sql  ---  although  the presenter I  asked  about  this  said  he

  recognized the requirement and  the  "mind  set" (like mine) that

  wanted a clean C api.

 

  A few interesting things  about  CM  (communications manager).  A

  new  feature,    called  LOC  (Lan  Over  Coax)  is a very clever

  feature.      As a result of  embracing  NDIS  (Microsoft/3comm's

  Network Device Interface Standard),   CM/LS  have  gained  a lot.

  This new LOC  feature  allows  you  to  run a LAN using your 3270

  connection card,  over standard coax cabling.  It is not  as fast

  as  a 4mbps token ring,  but it is very good news to people  that

  have a lot of this cable buried in the walls.

 

  CM  now comes with a real ASCII terminal program.  It is a subset

  of  Softerm  Modular,   from Softronics.  It has a whole bunch of

  terminal emulation modes and file transfer protocols,   including

  IND$FILE (zmodem to arrive  in  late  91).   Softronics has look-

  alike programs for DOS/Windows.  It is not a PM program.

 

  It looks like we are  still  stuck  with  C:\MUGLIB,    so  it is

  possible that MS Lan Manager will not coexist with ES.

 

  Network Services/2

  ------------------

  This  is  a new product that will  probably  be  merged  into  ES

  (Extended services).  It may  even  be  part  of  ES  when  it is

  released later this year.  However,  NS/2 is  available  now  (it

  was  just announced recently) for about us$200  (for  EE  1.2  or

  1.3).

 

  NS/2 is a complete rewrite of APPC,  and much more.

 

  Bob Trower and I bumped into the presenter,  Steve Joyce,  in the

  demo room the day before his presentation.   As  soon  as he said

  "I've  got  a better APPC",  an alarm went off in  my  head  (Run

  away!    Run  Away!  Run Away!),  and I ditched him as quick as I

  could.   Anyone who has used APPC under EE OS/2 knows how I feel.

  Just configuring it can turn your hair gray.  From  a programming

  viewpoint,    a  mature man can be brought to tears trying to get

  those first few bytes across the  lan.   Named pipes look so much

  better you'll consider  waiting  for them to implement them under

  TSO and/or VM <g>.

 

  After  Steve's  presentation,    I  had to go back and  apologize

  profusely.    Besides  encapsulating an exciting idea called APPN

  (Advanced Peer to Peer Networking),  NS/2  will install/configure

  over top of CM with three parameters --- you never have to wander

  aimlessly around the CM config screens.

 

  APPN  is  a  concept  where  you  don't  have to  rigidly  define

  everything.  It is part of SNA/SAA.  It dynamically finds  LUs on

  the  network  and picks the best route,    rather  than  the  way

  something like VTAM works today.

 

  NS/2  also  has  a new APPC programming  interface  called  CPI-C

  (Common programming interface),   which  supplies  a  100% source

  compatible  APPC  interface   across  all  SAA  platforms  (3090,

  AS/400,etc).   CPI-C is still  not  as  nice as Named Pipes,  but

  Steve  said  they  were  working  in  an   Open/Close  Read/Write

  interface that  most  micro  programmers  can  wrap  their brains

  around.

 

  APPC has DRASTICALLY improved performance (double),  with reduced

  CPU  usage (50%).  All current OS/2 APPC  apps  gain  from  these

  benefits.    Faster  than Named pipes.  Sends at about 7mbps on a

  16mpbs TRN (under os/2 1.3).

 

  NS/2 comes with REAL sample programs.

 

  Besides the simple config (you don't even have to define  the TPs

  if you don't want to!) NS/2 makes the APPC programmer's life much

  easier with better  error  codes.    Before,    if  your ALLOCATE

  failed,  you weren't told WHY.  (It could have been any number of

  things,  like  a  screwed  up  CM config,  bad wiring, programmer

  error,  etc).  NS/2 will tell you why your ALLOCATE failed.  This

  one little  feature  can  potentially  save  days  of programming

  hours.

 

  NS/2   has  "local-remote  transparency".    This  improves   its

  performance,    but  more  importantly,   lets you write an  APPC

  program with only one computer (one machine debug).

 

  NS/2 comes with some  utilities  that you may have seen before as

  IBM IUO (Internal  Use  Only)  utilities,   such as an APPC TRACE

  file analyzer.

 

  I don't plan on rushing out and writing  a  bunch  of  APPC apps,

  but you can tell I liked  NS/2.   I'll perfectly happy with Named

  Pipes most of the time,  and as long as the app is running solely

  on  a lan (no host interaction) I'll continue to use them.  But I

  think NS/2 has drastically  removed  my  fear of programming APPC

  again --- especially when the host programmers and I will be able

  to talk the same language (CPI-C).

 

  NDIS

  ----

  They went on and on about this.  It was not why I went to an OS/2

  seminar.  If I plan on writing an NDIS driver,  I'll download the

  spec and read it,  thank you very much.

 

  It was nice to know that 100mbps token ring lans are on their way

  though.

 

  Further to a  Ethernet/Token  ring argument I've been having with

  some people,  IBM  had  a good real-world example where TR shines

  over  Ethernet.  In their full motion video  (DVI)  work,    they

  found that they could only  have one station running video across

  a  10mbps  Ethernet,   since Ethernet is  a  collision-based  bus

  network.  If they added another workstation,  the video and audio

  would go  out  of  sync,    or  become too choppy to watch.  A 15

  station TR (16mbps) lan had no problem running the same s/w.

 

  Lan Server (LS) 2.00

  --------------------

 

  Almost the same as Lan Manager 2.00 <g>.

 

  The  largest  differences  are in the user  interface  (not  even

  remotely similar) and the fact that LS 2.00 does  NOT  HAVE  PEER

  SERVICES.  (You should have seen me standing there with  my mouth

  open  in disbelief).   They did not even mention it --- I had  to

  ask them the question point blank during Q&A.   They said the way

  MS  implemented  peer  services  was  not  good  enough  for  IBM

  customers.  (So IBM customers get nothing --- go figure).

 

  For  those  who  don't  know,    Peer  Services  allows  an  OS/2

  workstation  to  become  a  limited  server.  It can server  disk

  resources (one user at a time),  or print  resources  (more  than

  one person at  a  time).  You can't do this with LS 2.00,  so you

  will have to buy the full  blown  server software just to share a

  printer.

 

  There are some good things to say  about  LS.    The  Net  API is

  allegedly 100% compatible with Lan Manager 2.00 now.

 

  Since it uses NDIS drivers,  dozens of third party  network cards

  will work with LS now.

 

  LS has HPFS386 (available with LM about a year ago).   This gives

  much better remote performance,   and has new features like Local

  Security.

 

  HPFS386, CACHE.EXE,   and the "ring 0 server" (netwksta.sys) have

  an  intimate  relationship.      When  a  request  comes  from  a

  workstation,    the  request has a good chance of being satisfied

  completely in ring 0.  This can drastically improves performance,

  because before it would have taken a round-about route up to ring

  3 (to the main server code)  and  than through the OS/2 kernel to

  the  file  system (at ring 0) and then back up to  ring  3,  etc.

  HPFS386 never even goes through the OS/2 kernel.     Since  local

  applications  run  at  ring 3,  and go though the kernel for file

  system access,  running HPFS386 has no (or very little) effect on

  local application performance.

 

  Unfortunately,  this  makes  LS 2.00 incompatible with OS/2 2.00!

  After drilling the speaker about it,  she finally agreed  that if

  I used the default HPFS.IFS that  ships  with V2,  there would be

  a  good chance that it will work;  but I'll have to wait till  LS

  2.00 is released to know for sure.

 

  MS shipped a mutli-processor version of Lan Man to Compaq,  which

  will run on their SystemPro (I'm not sure if it is released yet).

  IBM's multi-processor  version looks better.  Not only is HPFS386

  on the second CPU (as with  the  Compaq  version),  but so is the

  ring  0 server, netbois and the NDIS drivers.    Coupled  with  a

  busmaster Token Ring  card,    it  is possible that a remote data

  request could be satisfied  without  ever using the main cpu.  If

  the data is located on disk (not cached),  then the main CPU will

  be  needed.    However,    since  Microchanel  supports  peer-bus

  masters,  it would be possible  for  the TR card to talk directly

  to something like a bus master SCSI card for the data --- this is

  in the future though.  (is  this  sort  of thing even in the EISA

  spec?).

 

  LS 2.00 comes  with  all of the disk mirroring/duplexing software

  than LM 2.00 has.  UPS support too.

 

  One unique feature to LS is "dynamic sharing".  A server does not

  SHARE a device until the first user tries to USE it.

 

  LS will run on OS/2 SE,  including OEM OS/2 and hardware (Compaq,

  Olivetti, Siemens, Tandy and Zenith were given as examples).

 

  RIPL (Remote Initial Program Load) is supported for  OS/2.   Each

  RIPL client can have a separate subset of files (config.sys,  for

  example).

 

  Novel support to be released in  3  phases.  Even after the final

  phase,  netware clients  will  not  have  access to LS/LM servers

  (unless they load two requestors).

 

  IBM is setting up  a  LS  application certification program (like

  Novel has,  except that IBMs is free).

 

  A busmaster TR driver will be available for OS/2.    A  benchtest

  showed that with this driver the CPU utilization dropped from 70%

  to 15% on a particular test.

 

  SPM/2 (System Performance Monitor/2)

  ====================================

 

  This tool is available now in the US,  and on June 28 in Canada.

 

 

  SPM/2 is a combination  of  some separate tools we have seen from

  IBM and elsewhere.   It has a CPU Usage  Monitor,   Physical Disk

  Access meter (by  monitoring  the  disk  DD),  and a Memory Usage

  Monitor.  These three  run  as separate child windows in the same

  PM app.

 

  The  last  one  looked the most interesting (since we've all seen

  many CPU monitors).   The  Memory  monitor shows many things in a

  combined graph,  so you can clearly see the relationship each has

  with each other.  For example,  the system  will  start  swapping

  the  moment the "working set" of  memory  exceeds  your  physical

  storage.

 

  SPM/2 receives  its  data  via  a  named  pipe from the DCF (Data

  Collection Facility).  Since it uses a named pipe,   the computer

  you are monitoring can be a remote computer.  The pieces are sold

  separately (SPM/2 with DCF is us$245,  and DCF alone is $75).

 

  THESEUS is included too (They plan to clean up the user interface

  of it soon).

 

  The data  can  also  be  dumped to a log or spreadsheet file,  so

  that a remote operator (using something like Netview/DM) can grab

  the data off of the remote servers and analyze it centrally.

 

  The named pipe data  is  completely documented;  so you can write

  your own real-time apps using the data if you want.

 

  Since the Disk Activity monitor has a close relationship with the

  disk device driver (disk0?.sys/base0?.sys),  this feature may not

  work with OEM versions of OS/2,    if  the OEM changes the DD and

  does not leave the proper hooks  in.    IBM OS/2 1.2 and 1.3 have

  these hooks today;  someone like Compaq may not have them.

 

  Multimedia (MM)

  ---------------

 

  IBM will be shipping beta OS/2 Multimedia extensions this year.

 

  IBM does not see DOS as a good environment for MM,    so  it will

  not be supported very much.  They will be concentrating on OS/2.

 

  Since 1991 seems to be the  year  everyone is defining MM,  there

  is  not too much to tell you.  The demonstrations IBM did for use

  were impressive though.

 

  One thing that caught  my  attention was a 16bit ACPA card (Audio

  Capture and  Playback  Adapter)  which  will  be  compatible with

  SoundBlaster cards.  (Isn't  Compact Disk technology  based  on a

  16bit DAC?  This card will sound awesome).

 

  MISC

  ----

 

  OS/2 1.3 contains 1500 source code changes to fix problems in 1.2

  (the last 1.2 CSD fixed 1000 of them).

 

  IBM is working with the following companies to test ES and/or V2:

 

 

       AST      Siemens  Acer          Everex

       Netframe CompuAdd Hewlett       Packard

       Dell     Olivetti Tandy/Grid    Compaq

       AT&T     Epson    Tandon        Reply

       Trichord NEC      NCR           Apricot

 

  Some new toys were recently announced by IBM:

 

       - 128MB optical diskette.

       - 8516 Touch screen.    This  monitor  uses a new technology

       called a "Strain Gage" which is built right into the monitor

       glass.  If you look at  an  8516,   it looks EXACTLY like an

       8515.     The  strain  gage  is  NOT based on capacitance or

       resistance,   therefore you can press  on  the  screen  with

       anything you want (pen, gloves,  hammer <g>, etc),  not just

       a human finger.  There are 256 pressure levels.

 



Comments

Popular posts from this blog

BOTTOM LIVE script

Evidence supporting quantum information processing in animals

ARMIES OF CHAOS