chen@digital.sps.mot.com (Jinfu Chen) (05/11/91)
Following are *paraphrased* excerpts from the paper titled "Domain/OS and HP
OSF/1: Interoperability and Operating System Evolution", dated Feb 14, 1991.
It is a 33 pages paper with lots of valuable information along with marketing
hypes. I believe HP made a similar presentation on the ADUS SysAdmin
Conference.
Disclaimer: I select the portion that I think relavant to the readers of the
newsgroups, omission is purely to save my time to type. I claim full
responsibility for typos, and mis-representation of information from the
paper due to paraphrase (English is not my native language). I suggest anyone
who is seriously interested in the information should obtain a copy of the
paper directly from HP/Apollo. I will not attemp to answer any question
regarding HP's plan. I hope HP people can provide us more informal
information. Finally, I speak for myself and my employer is not even a member
of OSF (I think Moto is a member of UI).
My personal comments are enclosed by {}.
Basically HPollo is planning a two phases migration: First phase is SR10.4
along with OSF/1 1.a on the Series 700, the second phase is SR10.x along with
OSF/1 1.b on all 68K platforms.
1991 1992
Q2 Q3 Q4 Q1 Q2 Q3
Domain/OS Releases schedule:
PSK: SR10.4 SR10.x
X11R4 Standards DCE
Motif 1.1 Full NFS
CDROM HP VUE 2.0
DAT Additional Disks
JAWS
DN10K X/Motif clients
HP OSF/1 Releases schedule:
OSF/1 1.a OSF/1 1.b
Full features for Add Support for:
End Users and VABS
Series 700 Series 400 (2D)
2D & 3D DN with 040 (2D)
CASE Tools DCE
Many Domain Features Multi-byte chars
Full NFS
More details about the four releases (10.4, 10.x, and OSF/1 1.a, 1.b):
SR10.4:
Standards:
10.4 will be fully compliant with XPG3, POSIX, OSF OSC AES (OSF Operating
System Component Application Environment Specification), FIPS, and SVID.
Full NFS:
In addition to current NFS support, Yellow Pages, Lock Manager,
Auto-mounter, PC-NFS, and user interfaces like XDR, Secure RPC, REX, RPC
GEN, and NFS API will be added.
User Interfaces:
Full X11R4 release on 68K platforms, and X11R4 clients on PRISM (sans
server). HP VUE 2.0 will be available.
{ I think HP is pushing HP VUE as *the* window manager which is okay to me
privided the DM Editor promised (see below) is delivered.}
Hardware Enhancements:
the Magneto-Optical Disk Autochanger (JAWS) will be supported on the
Series 400 systems.
Both the 400 MB SCSI and 1.4 GB SCSI disk will be supported by Domain/OS
at or about the same time as similar support under HP-UX.
SR10.x {SR10.5 ?}:
DCE will be implemented for Domain/OS.DCE will provide a common registry
system for user identification and validation, a single common file
system with single protection scheme (ACLS), and a single distributed
computing mechanmism (2.0) within a network of Domain/OS and OSF/1
systems.
OSF/1 1.a:
Platforms:
Only on Series 700.
Graphics:
X11R4, 2D and 3D Starbase, PHIGS+ and GKS.
CASE Tools:
DDE and DPAK will be available. HP LINT and GPR-to-X emulator tools will
aid Domain/OS users in changing their software to use standards-based
interfaces. The DCE toolkit will allow users to develop powerful
distributed applications for HP OSF/1.
Networking:
OSF/1 1.a based system will tie into existing network with TCP/IP and
NFS. Both the Apollo Token Ring and Ethernet physical links will be
supported. { No rewiring is required for existing ATR network, confirmed
in ADUS SysAdmin Conf. }
System Administration:
Three types of technique for user account administration will be
provided the manual editing of /etc/passwd files, NFS Yellow Pages, or
Domain/OS registry clients. These techniques do not work in combination.
Rather, one of three is chosen and used exclusively. The server for the
Domain/OS registry client must reside on a Domain/OS workstation on the
network.
OmniBacck will be available.
HP OSF/1 software will be distributed on CD ROM only.
OSF/1 1.b:
Platforms:
All 040 based workstations.
Graphics:
X11R4 on all 2D graphics controllers on the Series 400 (MVRX, CVRX, and
CRX). The DN5500 will support X11R4 on the M, E, F options and DVS
controllers. Later in 1992, on the Series 400 only, will be 2D and 3D
Starbase, HP PHIGS, and GKS.
Networking:
DCE will provide many features of Domain networking, including a global
filesystem, a global login facility, extended ACLs, and file lock.
System Administration:
Networks of Domain/OS, HP-UX, and HP OSF/1 can be all be administrated
in one simple consistent way with features provided by DCE.
Interationalization:
Multi-byte character sets will be supported in the XPG/4 style.
Interoperability in two phases.
Phase 1 (Q4 1991):
Networking - Domain/NFS
Full support of NFS with 2 to 4 times faster speed.
System Administration
Common tools available on both Domain/OS SR10.4 and OSF/1 1.a:
OminBack
Net LS
Passwd Etc.
CDROM
NFS: Yellow Pages and Automounter
User Environment
X11R4, Motif 1.1 and HP VUE2.0 on both SR10.4 and OSF/1 1.a.
DCE Tool Set
Available on Series 700 for software developers.
CASE Tools - DSEE, DDE, DPAK
DSEE can be used on Domain/OS workstations to build binaries targeting
OSF/1. Native DSEE on OSF/1 will be provided in the future (second half
of 1992).
DSEE 4.0 will have Motif interface.
DDE and DPAK will be available in HP OSF/1 1.a.
A Lint toll with user selectable database is planned to assist customers
in transitioning their applications. For example, the Lint tool will
check Domain/OS input sources for Aegis calls, then refer to the
appropriate database for the suggested XPG3 call. {I wonder what Unix
call it will refer for mbx_$ and ec2_$ calls}
The following is the list of the non-graphic database planned (the
graphic database will be described in the next section):
Aegis to XPG3
Berkeley to XPG3
Sys V to XPG3
SOCKETS to Streams
Remaining OS Calls to Standards
K & R C to ANSI C
Graphics - 2D and 3D
For the Domain user, the OSF/1 2D and 3D graphics capabilities, released
in fourth quarter of 1991, require that applications to be written in X,
PHIGS+, or Starbase. The following capabilities are offered to help
convert existing graphics:
running natively:
o Lint Tool Database for Graphics
o GPR to X emulation program
o Domain X to OSF/1 X recompile
o Domain PHIGS to HP PHIGS recompile
The Lint Tool will be enchanced with databases to convert graphic calls.
Specifically:
GPR to X
GMR to PHIGS
GPR to Starbase
GMR to Starbase
With the GPR to X emulation program simple GPR output calls are
translated into X calls. No GPR input is supported. There is a
significant penalty for using this program, but simple programs using GPR
output primitives can be emulated. Sources for this program will be
distributed.
User Interface - Cut and Paste Visual Editor
An OSF/1-based editor along with OSF/Motif modifications (startup files
and macros) that provides a *DM look and feel* {in bold letter
originally} are {sic} planned. The features planned for this editor
include:
o Third mouse key that opens a file for reading/editing.
o Key mapped to DM functions, where possible.
o Cut and paste between windows, files, and into the editor.
o A WYSIWYG editor that mimics the DM editor, as much as possible.
o Appropriate startup files and macros that implement this
functionality.
Phase 2 (Q2 1992):
DCE vs. Domain File System
While the global name space is preserved in the DCE, there is a
fundamental change in the model that needs to be understood. Traditional
Domain networks exist under a peer-to-peer model, where every machine is
both a client and a server. DCE File System (DCE/DFS) networks operate
under a client-server model where there is a small set of designated
servers that make their file systems accessable to all, and a large
number of clients that typically use their local disk space for caching
remote files from the server. Note that under DCE/DFS clients can make
their file systems apparent to other systems within the cell, where a
cell is defined as a signle administrative entity.
With Domain DCE, the best file access efficiencies are handled by DCE.
Decisions concerning locating files on servers or on the user's local
disk are made by DCE, base on file usage. This entire operation is
transparent to the user and automatically implemented within DCE. Domain
networking requires the user and system administrator to make these
decisions, and to reevaluate these decisions if file usage changes.
Client-to-server ratios need to be calculated based on the particular
applications environment for best performance.
The following is the DCE Domain/OS and OSF/1 support matrix.
OSF Domain/OS Description
DCE DCE
AFS 4.0 X X The distributed file system for OSF/1
provding uniform name space, file
locations transparency, and high
availability
Episode Physical
File System X + The physical file system for OSF/1
Naming (including X.500)X X OSF's X/Open compliant naming model
throughout the distributed
environment
Time X X The distributed time reference for
scheduling activities and determing
event sequencing and duration
Threads X X A facility that supports concurrent
programing
Security X X Services that are well integrated
within the fundamental OSF
distributed services and data-sharing
components
RPC X X Permits individual procedures in an
application to run on a computer
somewhere else in the network
PC-NFS X * Gives minicomputers, mainframes and
PC users the ability to share files,
peripherals, and applications in a
distributed environment
Diskless X # Accomodates diskless workstations and
provideds well defined, general-
purpose protocols for diskless
support
LM/X X @ HP's PC LAN Manager
+ Domain/OS has its own physical file system
* Domain/PC-NFS is planned with the enchanced NFS product
# Domain/OS supports its own diskless capability
@ Domain/OS will not support LM/X
+ and # are due to continue support of Domain-to-Domain services. The
Domain/PC Interconnect product offers many of the same features as LM/X.
A Domain/OS unique file format, i.e. typed files, which is used
extensively in DSEE is not accessible through DCE. However, Domain DSEE
Version 4.0 can be used to manage OSF/1 binaries and source files. HP is
claiming their own OSF developements use Domain DSEE to manage the
binaries {at Chemsford or at Ft. Collins?}
OSF/1 nodes wil not operate as Domain network protocol routers, and thus
access from Domain/OS nodes to Domain/OS nodes through an OSF/DCE or HP
OSF/1 nodes using the Domain network protocols is not possible. Routers
to handle Domain network protocols are available from HP and cisco.
SR9.7 and SR10.[0-3] nodes will not interoperate with OSF/DCE or HP OSF/1
nodes, except through the Domain/NFS product. However, SR10.4 and greater
nodes will interoperate with DCE.
Other Networking
To support wide area DCE and IBM communications the following networking
services are planned for 1992 delivery:
o X.25
o ISO/OSI
o SNA 3270/3770
Commonly Used Domainisms
The following features of Domain/OS represent some of those that are
highly valued by Domain users. Along with each Domain/OS features is an
explanation of the identical or similar feature provided by HP OSF/1 or
DCE.
o Network wide file system: The AFS provides a single network wide file
system, very similar to that of Domain/OS. But where the Domain/OS
file system is only available in local area networks, the AFS offers
a network wide file system that works over wide area networks as
well.
o Concurrency control: The AFS provides similar control to Domain/OS.
o Global location broker: DCE includes NCS which includes GLB.
o Global login: The Passwd ETC. part of the DCE offers identical
capability to the Domain/OS registry.
o Cut/paste visual editor: There are a number of X-based visual editors
currently under consideration for inclusion within the HP OSF/1
operating system. All of them offer similar features to those found
in the DM editor in the Domain/OS, including cut and paste
capability. These X-based editors will be enchanced to be more like
the DM editor.
o DM infinite undo: Although mwm offers undo capability, it does not
keep an infinite history. The infinite undo capability could be added
to mwm at a later date.
o CRP: In a "pure" UNIX environment, remote login and remote execution
are possible with the standard TCP/IP utilities of rlogin and rsh.
o Kep mapping: With mwm there are startup files to define key mapping.
o Input/output pads: The X window system and mwm do not currently
support the concept of pads backed by data files on disk. The windows
managed by the mwm are backed by buffers in memory and have a limited
size (that is configured by user startup files). A window manager
that uses files on disk to back up windows could be developed at a
later data. { That is not what we want, we'd like the stacking
behavior of (input) pad }
o Popular Domain commands: There are a few Domain commands that are
very powerful and familar to the Domain users. As an aid to
migration, many of these services could be provided on OSF as
portable user-space servers with little extra effort. This is a
matter for further investigation. {like lcnode? or /bin/ps //node?}
o Domain mailboxes: Codes that uses mailboxes will have the choice of
using NCS RPC or Socket calls in their OSF/1 implementation.