New or updated items are marked with italic
This FAQ was created by Martin Alfredsson and
was edited (English) by Daniel McNulty.
Please note that most features/timetables discussed are rumors and may never surface.
"WorkSpace OS"
Well, WorkSpace On Demand (AKA Bluebird) is now official and it brings me back to
the rumors of a change of name. Do you remember the WorkPlace OS, the OS that lead
up to OS/2 for the PowerPC ? Let's make a qualified guess: merge these names. Well
how about WorkSpace OS or WorkSpace OS/2 (for a transitional period).
I can already picture the ad: "WorkSpace OS - Your future WorkSpace, built
on proven OS/2 technology".
I for one would not care if they changed the name. Just keep WorkSpace Shell (deliberate
spelling mistake) and the OS/2 API and I'll love to install WorkSpace OS Version
5.0 on my machine.
Netscape Navigator 4.0
What does this have to do with Warp 5?
Many sources tells me that the reason for the delay in the project (or at least
one of the reasons) is that IBM want the Navigator Part of Communicator as the primary
browser for OS/2. This means the final end of WebExplorer and more importantly that
IBM gets the source code of Navigator. IBM has a policy that states that all software
included in their operating systems must be owned (or at least part-owned) by IBM
and that IBM must have the source code to fix bugs and so on. I'll make a qualified
guess of the browser that's part of Warp 5: Netscape Navigator 4.0.
Warp Server 5.0 in Alpha
Warp Server 5.0 is rumored to go into alpha testing (outside of IBM) in a few weeks.
I don't know for sure but I believe this is Warp Server 5.0 (LAN Server SMP atop
Warp 4) and not Warp 5.0 Server (the server version of Warp 5). If anyone knows
more please inform me.
Warp 5 - Very unofficial :-)
I spoke with a IBM developer from Raleigh, NC. I did ask him about the Warp 5 beta.
He stated that he had seen it and confirmed that it was being worked on actively.
He also confirmed the merging of the source code for the server and desktop version.
He offered testing betas when they became available.
I also asked some other highly placed IBM employee about specifics and received
some very wide grins with shuffling of feet...very unofficial.
No more Warp 4.1 rumors
Though I don't doubt that there will be some kind of "service update"
or "fixpack+" released for Warp 4 I do not believe there will be a release
of Warp that is called Warp 4.1. Therefore I have removed the section on Warp 4.1
in this FAQ. I will however report on related subjects on my main
page.
JFS/Warp on internal IBM disks
This is a directory listing I got before summer:
jfs.ifs 28-May-97 15:01 193k
jfs.sym 28-May-97 15:01 12k
jfssyncd.exe 28-May-97 15:02 29k
jfssyncd.sym 28-May-97 15:02 3k
logredo.exe 28-May-97 15:02 65k
logredo.sym 28-May-97 15:02 4k
mkfs.exe 28-May-97 15:02 63k
mkfs.sym 28-May-97 15:02 4k
ujfs.dll 28-May-97 15:02 9k
ujfs.sym 28-May-97 15:02 2k
(And, no, I do not have to code myself.)
So you still believe that Warp 4 was the last version of OS/2? Get a grip and get
ready for a reality check. IBM has (yes IBM) officially announced the next version
of OS/2 will be released in early 1998. They refer to it as the "the mega release".
This document refers to personnel requirements and also refers to Warp 5 as Warp
5/Merlin 2. According to this document there are only a few features that are firm
at this time. The decisions on the final feature set will be made in October/November
1997.
Apparently there are several different teams evaluating several different plans
(directions) for Warp 5/Merlin 2. This means that we will probably hear lots of
contradictory rumors until IBM decides what path to follow. Some people will tell
you that Warp 5 will NOT be based on this or that and other people will tell you
IT WILL. Not even IBM will know for sure until October/November 1997!
Warp 5/Merlin 2 has over a dozen (IBM internal) system design documents (SDDS) that
will be available over the next three months that describes the function being provided
by a Warp 5/Merlin 2 system (device drivers, file systems, system management, installation,
transports etc.), the components that constitute the system, and the interaction
of functions and components.
Each of these SDDs will, in turn, spawn one or more (IBM internal) component design
documents (CDDs) which provide the detailed design for each component. The development
of the CDDs will overlap that of the SDDs and is targeted to continue through November.
If anyone has any of these SDDs or CDDs please send them to me. The best way
to stay anonymous is to scan the pages on a scanner and e-mail the bit maps to me
or even OCR the pages. Mail me: jma@jmast.se
One very interesting thing is the talk of "significant OEM emphasis for Warp
5/Merlin 2". This seems to mean that IBM will try to make Warp 5 an even better
network client. That is a LAN client, not just Java!
The release date is also a moving target (software development folks!). The plan
at PSP is to release in the first quarter of 1998. Warp 5 it is to be client/server
and it means that IBM creates a single version that is used for both Warp Server
5 and Warp Client 5. That will make things much easier for both IBM and users (one
fixpak for both versions).
Another SDD (system design document) for the Warp 5 kernel that I have seen though
talk about the design of the Warp 5 kernel. However this document was last edited
in late 1996.
Will it be Microkernel based ?
Probably not, there has been no indication since early this year of any OS/2 Mk
development. If I'm wrong, tell me, I will not be sorry.
No more 512MB limit
"Warp 5 will provide all 32 bit applications the ability to address memory
higher than 512 Mb. Applications that use APIs that accept pointers as parameters
can now allocate those parameters above the 512M line." Note: This is a requirement
if, and I say if, IBM wishes to support Win32s 1.30
Kernel redesign
"Another goal is to make all the internal worker routines pure 32 bit (enhancing
performance). The external interfaces to the Doscall APIs will not be changed to
preserve compatibility. Thunk routines will be removed and new interfaces will be
created in the Kernel to support 32 bit workers but external 16 bit APIs will still
be supported."
SMP everywhere
"Symmetric multiprocessing (SMP) support is a key feature in Warp 5. SMP support
will be fully integrated into Warp 5. The Warp 5 product will integrate SMP support
into a 'single source' product code base. There will not be a separate SMP product
for Warp 5. All versions of Warp 5, both client and server versions, will support
SMPs. The primary objectives of the Warp 5 SMP support is to integrate SMP support
as a basic part of the system and to improve the SMP scalability and performance.
In order to have both the cost saving attributable to a single binary and still
have acceptable Uni-processor (UP) performance, Warp 5 will have single SMP enabled
binaries for all parts of the system except the kernel (OS2KRNL). Both UP and SMP
versions of OS2KRNL will be provided to avoid the SMP performance cost when running
on a UP system. "
Time to get rid of your 386
"Warp 5 intends to use the CMPXCHG instruction. This instruction was introduced
on the Intel 486 processor, therefore Warp 5 will not work on Intel 386-based machines.
The usage of this instruction by Warp 5 on both UP and SMP systems prevents Warp
5 from running on Intel 386 processors. Some of the places where the CMPXCHG instruction
will be used is in the 32 bit RAM semaphore code and the kernel execution environment
locking services."
New device drivers
"Warp 5 assumes that all 32 bit installable file systems and device drivers
are SMP safe. The 32 bit IFSs and DDs are new for Warp 5. Warp 5 will not do protective
locking around 32 bit IFSs and DDs to make them SMP safe. The new 32 bit IFSs and
DDs must make themselves SMP safe. Kernel execution environment 32 bit locking services
are provided for use by IFSs and DDs to become SMP safe. Kernel execution environment
services are also provided to obtain necessary serialization for a 32 bit IFS or
DD to call a non-SMP safe 16 bit DD. " Note: This is the Add32 interface. I
think IBM will push developers hard to convert their 16bit device drivers to 32bit
versions telling them that "a future version of Warp will not work well with
16bit drivers".
Osprey
The code name Osprey (to prey on other operating systems) is the effort to supply
XPG4v2 (spec1170) and X11/Motif API's for Warp 5. This allows most UNIX (spec1170)
programs to be ported to OS/2 quite easily.
JFS/Warp
JFS/WARP is the OS/2 version of the AIX 64bit Journaling File System. It is currently
being ported to OS/2 by mainly a group of AIX people. JFS/Warp has been out on a
internal alpha/beta round a least since the beginning of February and the current
code is said to be quite stable. It is therefore possible that it will be out before
Warp
5 as a "fix" for Warp Server. IBM is VERY quiet about JFS/Warp and might
plan to use it as a surprise item when promoting some future Warp update (most likely
server based). I would expect IBM to deny its existence up till the day they tell
everybody what a new and fantastic file system they have for OS/2 Warp.
The code name for the JFS is "Owl" and here is why :-) The JFS/WARP team comes largely from the AIX Programming Center. The owl is their mascot, as UNIX programmers are often nocturnal creatures. UNIX is also the operating system of choice in academia, and owls are usually represented as being wise.
LVM
Logical Volume Manager (LVM). All Installable File Systems (IFS) will be contained
in the LVM. LVM takes over drive letter mapping for programs that require them.
This allows users to selectively assign drive letters to a volume of our choice.
And (even better) the OS itself won't use drive letters. I'm told that this combined
with JFS/Warp will allow us to combine several partitions and even physical disks
into a single logical drive (also known as disk spanning)!
OO drivers
There might also be 32 bit Object Oriented "device drivers". The technique
of Object Oriented device drivers is inherited from OS/2 for the PowerPC and is
probably based on Taligent code. It means that a device driver developer can "inherit"
from other drivers to greatly reduce the development effort. This is similar but
not that same as the current video (GRADD) and printer (OMNI) drivers and NOT the
same as ADD32 drivers.
More SmartGuides
The SmartGuides (WarpGuides in Warp 4) will be enhanced and expanded. There are
already a dozen tasks that are identified as candidates for Warp 5/Merlin 2. I suspect
there will be lots of work with Guides for the installation program.
Human-centric features.
Full human-centric features. What we see today is primarily VoiceType in Warp 4.
VoiceType is included in OS/2 for the PowerPC since the PowerPC has DSP encoding/decoding
in the hardware. We might see other things like voice output (text to speech), software
agents and an operating system that adapts to the user. There will probably also
be a more adaptable/customizable user interface. There are rumors of a much more
guided interface.
More horsepower
Warp Server will be much enhanced with new multithreaded device drivers (probably
from OS/2 for the PowerPC), no more serialized I/O, raw file I/O, full failover-support
and lots more. Clustering of servers will also be supported.
JFS/Warp
Many things point to a new file system being included into the next Warp Server.
The JFS/Warp file system is described above in the Warp 5 text but is also planned
to be included into Warp Server. The texts says that JFS/Warp is to be included
into the "Warp Server framework". I have no indications what this means.
Fallback Fault-tolerant Server
A server-specific enhancement called Fallback Fault-tolerant Server (FFS) will be
included. FFS can be thought of as an automated version of the domain controller/backup
server concept in Warp Server. The FFS is the same thing as the Vinca Standby server
as IBM licensed it from Vinca. And the concept is that there are 2 machines (servers)
up and running. One of them (the primary machine) will be running all the server
code, while the other ( the standby machine ) is only running OS/2. As soon as the
primary goes down the server code gets started on the standby machine which becomes
the primary machine. It takes from one to two minutes for the switch over.
In Austin there is a secure server with a directory that had a list of many of the
projects IBM PSP is working on. Each project has a directory and are password protected.
Well, there are two directories for Warp, one for Warp 5 and another one for Warp
6. Sorry to say no one who has been able to peek inside has yet contacted me.
If Warp 5 is non-Microkernel based you MK freaks could possibly get your hope up
for this one.
"JFS/WARP Project Overview"
"The Journaled File System for WARP is intended to satisfy the requirements
for a robust, competitive, byte-level file system in the WARP environment. A journaled
file system provides improved structural consistency and integrity using techniques
developed for database management systems. It logs information on the file system
metadata and is able to replay this log to commit or abort in-flight transactions
upon recovery from a system failure. Applying changes from the log not only ensures
consistency, it also markedly speeds recovery time over methods such as the chkdsk
command in OS/2 or the fschk command in UNIX. Although preserving the metadata,
it does not guarantee recovery of user data in the files."
"JFS for WARP will incorporate the POSIX and XPG4v2 (Spec 1170) file system
interfaces and semantics as its base set of functionality. This is in support of
the Osprey effort. The JFS for WARP will operate as a physical file system under
the WARP File Server Framework. It is based on the field-proven technology developed
for the AIX Journaled File System."
Even more on JFS/Warp
"I've seen an old draft of the spec for JFS for OS/2."
Some of the design criteria are:
1) As fast or faster than JFS under AIX (it screams)
2) As fast or faster than other JFS implementations under UNIX
3) As fast or faster than NTFS, or HPFS386
4) Very fast recovery from logical file system problems, even for very large disks
(unless there is physical damage to the disk or to the underlying structure; then
recovery time could be lengthy)
5) Very reliable, due the journaling feature
6) Maximum partition size of 2 petabytes using the maximum cluster size of 4096
bytes (I'm not even sure what a petabyte is!). This is likely to be artificially
limited to some smaller size, since there is no way to test a 2
petabyte partition. With a cluster size of 512 bytes, the maximum partition size
is 28 terabytes (I think).
7) The maximum file size is equal to the maximum file system size; this would also
be limited due to other constraints within the OS; the document did not state what
that size would be, but is might be the current limit of 2 gigabytes, unless some
other unspecified changes to the OS are made.
8) Supports striping and spanning
9) Is efficiently scalable for systems running up to eight processors
10) Will be disk image compatible through some mechanism with existing HPFS drives
"The document made constant references to interfacing to a microkernel, and
constant references to ADD32.
"Not sure what to make of that."