[刦K] The 86open FAQ
cosmos
順便貼 86open 的 FAQ 讓大家看看....
URL: http://www.telly.org/86open/86open-faq.html
---------------------------------------------------------------------------
The 86open FAQ
Last updated: December 23, 1997
This file contains a list of Frequently Asked Questions (hence the term
FAQ) regarding the 86open group, as well as answers to those questions. It
is written, maintained and Copyright (C)1997 by Evan Leibovitch. It may
be freely redistributed in unmodified form, and its primary source is
http://www.telly.org/86open/86open-faq.html though mirrors are
welcomed.
Notes:
The term Unix-on-Intel as used in this FAQ, refers to the subset of
Unix systems specifically produced to run on computers based on
industry-standard computer architectures using Intel(R) 80386 and
compatible CPUs.
Also, the term Unix, as used above and elsewhere in this FAQ,
describes the entire realm of both commercial UNIX(R) operating
systems and others such as Linux and BSD which have similar design.
PART 1 - HISTORY AND PURPOSE
Q1: What is 86open?
86open is a loose and informal group of technically-oriented people, who
work and/or program in the world of UNIX systems and compatibles on
Intel-based, standard-architecture computers (Unix-on-Intel)
Q2: What are 86open's goals?
Its goal is to address areas of common ground amongst a diverse group of
operating systems, including (but not limited to) multiple variants of
freeware OSs such as BSD and Linux, as well as commercial OSs from
BSDI, SCO and Sunsoft. Any agreements, specifications and software
produced by it will be made freely available (in source code where
applicable).
Q3: When did it start?
After some initial e-mail discussions took place, the first actual meeting
of the group took place at SCO's head office in Santa Cruz, California on
August 22, 1997. That was the first and only physical meeting, everything
since has taken place by e-mail.
Q4: Is it ongoing, or does it have a specific goal in mind, after attaining
which it would cease to exist?
This hasn't been discussed in great detail by the group. Work has already
started on 86open's first project, a library specification allowing a common
binary to run without emulation on all participating operating systems (see
Q15): While some participants have envisioned other projects once this
first one is complete, all attention is presently focused on the library
project and not beyond.
Q5: Isn't this just another one of the many Unix alliances?
No. 86open is significantly different from most previous Unix industry
groupings such as Unix International, the Open Software Foundation, the
COSE and ACE consortia, and others.
This group is made up of individuals, not companies. While many of
86open's participants share their own organizations' point of view,
few officially represent them. It is our belief that the quality of this
group's work, as well as its members' personal respect within their
communities, will enable its works to be implemented across the
Unix-on-Intel world.
Being a highly technical group, it members have a less-competitive
temperament than other groups which have been made up of
management or marketing representatives. Further, participation is
totally voluntary and nobody is here because they are forced to do so.
This is, of course, just opinion, but it appears to be borne out by many
members of the group.
86open is the first such group to have significant representation from
within the freeware (Linux and BSD) communities.
Q6: Aren't you just doing this as a reaction to Windows NT?
Yes and no. It would be a lie to deny that NT has provided a significant
challenge to the Unix community, and part of the weakness of the Unix
community is its high level of fragmentation. Anything that can be done to
reduce this degree of fragmentation, while allowing each participant to
maintain its unique set of strengths, is beneficial to the community as a
whole.
On the other hand, there are many who believe that, NT or not, this effort
is necessary and long overdue.
The participants in 86open generally believe that Unix-on-Intel provides
an unmatched combination of value and performance, and that its installed
base is large and growing rapidly. We hope that our efforts on joint
projects
will highlight these strengths before the computing public.
PART 2 - ORGANIZATION
Q7: Where and how is 86open incorporated?
86open is not incorporated, and there are no plans to formalize it to that
extent. It has no budget, all costs to date have been borne by individual
participants or their employers.
Q8: How is the group structured?
There are two main components to the 86open group:
The main group is about 30 people, who will jointly be involved in
producing projects mutually agreed-to. All decisions will be reached
by consensus of the group, with dissenting opinions accommodated
whenever possible.
A subset of this group, a seven-person 'steering committee', will
handle publicity and other administrative tasks. It will make no
technical decisions on its own.
To date the group has physically met once. Future meetings may be
convened but presently all collaborative work is done by e-mail.
Q9: What organizations are represented here?
It cannot be emphasized enough that 86open is comprised of individuals,
not organizations. While we are doing our best to ensure that all possible
viewpoints are represented, we work under the premise that nobody in this
group speaks on behalf of their parent organization unless they explicitly
say so.
Having said that, it is our desire that the fruits of our work are implemented
across the entire community of those who use and develop Unix-on-Intel.
To that end, we have tried to attract individuals who can offer perspectives
from across the community.
Current membership in the group reflects perspectives from Be Inc. BeOS,
BSDI BSD/OS, Caldera Linux, Debian GNU/Linux, FreeBSD, NetBSD,
Red Hat Linux, SCO(R) (OpenServer and UnixWare), Sunsoft Solaris,
Intel(R) and The Open Group.
Q10: Why isn't <my operating system> on that list?
We have made reasonable attempts to solicit participation that reflects
viewpoints from across the Unix-on-Intel spectrum.
We may have missed some, inadvertently. Developers of any relevant OS
are encouraged to mail us to help us fulfill our tasks. The only OS whose
developers have been invited to participate, but declined, is OpenBSD.
Q11: Who is involved?
Here is the current membership. Note that only names are given, not
affiliations, deliberately (see Q9). Members of the steering committee are
in bold:
Tim Bird, Keith Bostic, Chuck Cranor, Michael Davidson, Chris G.
Demetriou, Ulrich Drepper, Don Dugger, Marc Ewing, Steve Ginzburg, Jon
"maddog" Hall, Ron Holt, Jordan Hubbard, Dave Jensen, Dion Johnson,
Kean Johnston, Andrew Josey, Evan Leibovitch, Robert Lipe, Bela Lubkin,
Tim Marsland, Greg Page, Bruce Perens, Ron Record, Andrew Roach,
Tim Ruckle, Joel Silverstein, Bryan Sparks, Chia-pi Tien, Linus
Torvalds, Erik Troan
Q12: How can I become a member of the technical group?
If you can offer a viewpoint that we don't have at this point, and if you have
the time/resources/skills to work on projects in this group's plan, please
mail us. No promises -- there are concerns that if the group gets too large it
may be unworkable -- but we'll give all requests their due.
Q13: Is there any leader or chairperson?
At this time, no. Certain individuals have specific responsibilities, and so
far that's worked for us.
Q14: Why the name 86open?
It's simple, catchy, and describes both the group's focus (on Unix-on-Intel,
often referred to as "X86" architectures) and the fact that the products of
its work will be freely available.
It's also reminiscent of "88open", a respected industry group formed many
years ago to support the Motorola 88000 processor. The 88open group
ceased operations in 1994, and its last-known leadership is aware of our
group and has no objections to the use of the term 86open.
Q15: Do you have a logo?
We don't have one, and we're in no hurry to get one. Thanks for the offers,
but that's nowhere near the top of our priority list.
PART 3 - PROJECTS
Q16: So what exactly are you doing?
At the August meeting, the 86open group reached a consensus to put its
initial efforts into a method to allow a single binary program to run,
unmodified, on any participating OS. That is, one could develop a program
on OpenServer that would run on FreeBSD, or develop on Linux and run
on SolarisX86, and so on.
Q17: A single binary used by all Unix-on-Intel OSs -- who won?
This was not a battle to be won or lost. The 86open binary project is not an
emulation scheme in which one system's binary format becomes a standard
which others must support. Rather, a single binary, written to a standard
programming specification, will be able to run natively on all supporting
platforms.
Q18: How can you get a single binary to work identically across all these
diverse systems?
Most Unix-on-Intel binary packages are already largely similar. Almost all
such operating systems use the "ELF" binary 'packaging'; the various
operating systems have small but significant differences, though, that make
each system's ELF binary unusable on others'.
Because of earlier binary compatibility efforts such as the iBCS2 (which has
been a part of UNIX since System V/386 Release 3.2), most of the older
functionality is already quite similar between the systems. Most of the
divergence has been in post-iBCS2 developments such as lstat() and
mmap().
The 86open group plans to solve this problem by producing a standard set of
library functions that would be provided on each participating OS. A binary
using these standard functions would dynamically link into the local library
containing them, thus allowing it to operate regardless of the different
internals of the underlying OS.
It is up the library implementation on each OS to hide any OS-specific
behavior behind the standard functions, allowing programmers to
concentrate on a single Application Programming Interface (API) for a
binary that will work on all systems.
Q19: This might work in concept, but is it practical?
Yes, absolutely. One such proof-of-concept is the the lxrun project, which
allows Linux binaries to work on SCO and Solaris operating systems by
mapping Linux OS-specific system calls to their SCO equivalents by use of
a 'shim' library.
The feasibility of a common Unix-on-Intel binary by use of a single library
specification was agreed to by all the programmers involved.
Q20: What will be in this new library?
That's what's being worked on at this time. The emphasis is on defining a
set of functions that is "sufficient, yet limited", in the words of Linux
creator Linus Torvalds. The intention is to include all the basic functions
necessary to create applications (and act as a base for other libraries,
such as
for X Windows), while removing redundancies and unnecessary code
according to the groups' consensus.
The group is starting with a conventional Unix 'libc' (GNU's glibc2) as its
point of reference. While any participant is welcome to create its own
implementation of the 86open library specification, the reference will be
based on glibc.
Q21: What programs will benefit?
The intent of the 86open library specification is that it will be sufficiently
complete to allow most kinds of software applications to use it in lieu of the
OS-specific libraries required presently. We will attempt to offer
documentation to assist developers to port existing applications to the 86open
library functions. We understand that certain kinds of software that must
necessarily be close to the OS kernel will probably not benefit from the
portability aspects of the 86open specificationi. This is because such
programs, such as performance monitors, will require OS-specific facilities
beyond the scope of our efforts.
Q22: How will you ensure that the specification is open?
The license of GNU glibc requires that it's freely available in source code,
and that all modifications to it be similarly available. Furthermore, 86open
is committed to maintaining a reference implementation of its library
specification that is available in source code and freely re-distributable.
Q23: Won't this library specification create just another Unix 'standard'?
This is certainly not our intent. Wherever possible we will remain
consistent with The Open Group's Single Unix Specification. But there
may be areas the group needs to address --- such as direct video access ---
that are architecture-specific and thus outside the scope of the SUS.
Q24: When will the specifications and reference library be available?
Progress is being made. But it's too early to promise any deliverables, and
the consensus is that it's better to say "we're not sure" than to pick a date
out of thin air and lie. We like to think we're above that; being open can
mean more than releasing source code.
PART 4 - FUTURE
Q25: Is anything planned once the library specification is complete?
86open is fully aware that a common binary -- while a significant step -- is
only part of the solution to making the Unix-on-Intel platform attractive
and easy-to-support for application developers.
While nothing is concrete -- all attention is presently focused on the
library project -- further issues have been raised that may be later
addressed by 86open, such as installation systems and common filesystem
layouts. It must be emphasized, however, that future projects haven't been
dealt with any level of detail, and likely won't until the current one is
near completion.
SECTION 5 - LEGAL FODDER
Q26: Are there any trademarks in this FAQ that you need to identify?
Why, yes, glad you asked.
BeOS is a trademarks of Be, Inc.
BSD/OS and BSDI and are trademarks of Berkeley Software Design, Inc.
Intel is a registered trademark of Intel Corporation.
Linux is a trademark of Linus Torvalds.
Microsoft and Windows NT are either trademarks or registered trademarks
of Microsoft Corporation.
SCO, OpenServer and UnixWare are trademarks or registered trademarks
of the Santa Cruz Operation, Inc.
Solaris is a trademark of Sun Microsystems, Inc.
UNIX is a registered trademark of The Open Group.
There may be others. If we've left them out, it's accidental.
Q27: Is 86open a trademark?
Haven't given it any thought -- no lawyers in the group. Right now there's
no formal entity to 'own' it anyway.
For more information, email 86o...@telly.org or check out
http://www.telly.org/86open