------------------------------------------------------------------------------
 OD-90228                         SUGGESTION                       "PRIDE"-IF
------------------------------------------------------------------------------
 JUN 16, 1997                   RELEASE  1.2.0                        PAGE  1

       ****************************************************************
       *                    MI-10023 SIQ SOLUTION                     *
       ****************************************************************

 REQUIRED DATE- OPEN          USER PRIORITY WEIGHT-  51    RANKING-  0019
 DATE RECEIVED- APR 23, 1997  REQUEST STANDING  - REVIEW   AS OF- JUN 16, 1997
                              TYPE- MAINT

 DESCRIPTION-

        The Synchronous Input Queue is the single largest stumbling
        block in OS/2's famed stability.  Even running no 16-bit
        applications (that I'm aware of) I've had my system completely
        lockup on numerous occasions.  I'm running Warp 3 (red spine)
        with FixPak 26.  I understand Warp 4 is somewhat more stable,
        but that it is still quite capable of locking up due to the SIQ.
        What I suggest is a sort of "Virtual" SIQ whereby older apps
        (16-bit) can see the SIQ and thus function normally, but newer
        32-bit applications which do not require the SIQ can run in
        true multitasking, asynchronous mode.

        ATTRIBUTES:

        HIGH Value
        Desired NEXT full release

        Submitted by:

        Don Eitner
        Upland, CA, USA
        freiheit@tstonramp.com

        ADDITIONAL TEXT -

        In Warp 4, even with the FP1 applied, there still is a large
        tendency for the "SIQ fix" to not work.  In particular, this
        problem shows in two areas:  SEAMLESS Windows apps and Java-
        intensive web pages.  Replacing SEAMLESS.DLL works around the
        problem with seamless Windows, but Java apps still appear to
        have a memory leak.  All in all, the SIQ fix for Warp 3/
        Connect in FP17 worked a lot better.  I rarely have seen it
        work at all in Warp 4.  After a lot of NetScape browsing of
        pages with Java applets, the system comes to a crawl.  If
        noticed in time, NS/2 can be closed and the system will, after
        several minutes, return to normal.  If not noticed in time,
        there is no way that NS/2 or even the WPS can be closed.
        Even Ctrl+Alt+Del will fail.  No modern crash-protected
        operating system should allow this to happen.  It is a
        complete turn-off for novices to the operating system, and
        a source of frustration for even seasoned users.  Mission-
        critical machines using Java should beware.  Naturally,
        whatever the Java problem is should be fixed, but most can
        understand that this is still a new technology.  Operating
        systems, however, are not new technologies, and how the SIQ
        fix, which mostly worked in Warp 3, has changed, I do not
        know or understand why.

        Submitted by:

        John Carmack
        Broadview Height, OH, USA
        MarchHare@momeraths.org

        ADDITIONAL TEXT -

        PM/WPS should support multiple Input Queue, changes to
        environment variables without reboot and it should be much
        faster.  End users usually judge the stability and speed of
        the operating system by his GUI, and that is where NT's take
        the cream.  End users see in NT a faster and more stable OS
        than OS/2 because it doesn't have SIQ problems nor is as slow
        as PM.  What is also a problem is a needed reboot of OS/2 after
        installation of software.  But I noticed that software  usually
        just updates the environment variables like LIBPATH and PATH.
        Fixing this would bring another (big) plus to OS/2 and it
        shouldn't be hard to implement.

        Submitted by:

        Rok Papez
        Ljubljana, Slovenia
        rok.papez@kiss.uni-lj.si

                           * * * END OF OUTPUT * * *