Date: Mon, 8 Mar 1999 09:50:21 -0500
Reply-To: greg.barnesnelson@statprobe.com
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Greg Barnes Nelson <greg.barnesnelson@STATPROBE.COM>
Subject: Re: A version of SAS which runs under Linux?
Content-Type: multipart/alternative;
Tim, as always, you make some very compelling arguments for a freeSAS. I
also learned three new words every time I read one of your posts. thanks.
Because of SAS' predominance in server-based architectures, Linux provides a
nice, low-cost architecture for departmental and workgroup computing. The
TCO (total cost of ownership) however, doesn't just stop with the
architecture. You still need people to maintain the box, etc. An even more
compelling reason to have a low-cost OS. Various vendors (IBM, Dell,
Compaq, Sybase, Oracle) are moving to support this platform. SAS should
consider this as a viable OS, despite it's rough-edged appearance.
Some points to consider when thinking about a freeSAS.
1. Because of Linux's widespread use in Internet and Intranet applications,
I would hate to see a version releaseed without some of the more interesting
products such as SAS/IntrNet and MDDB server. (see Tim's comments below on
releasing a 6.0x version of freeSAS)
2. SAS has a reputation for world-class technical support. I doubt that
they would simply hand over the support to a group of well-intentioned,
knowledgeable, SAS-loving renagades - no matter how capable they might be,
or how small the market segment.
3. Reviving 6.04 support levels (even from a re-compile perspective), would
require a considerable amount of energy. Even when I was at the Univ of
Georgia and we had 6.02 users out there (because it ran well on 286/ 386
machines), we had to special-request setinits for the older versions. And
because it was not maintstream, it requires extra effort. The recompile
would be fairly tough considering the modern (current) changement
management/ version control tools that they use have most likely been
introduced since 6.11. Although I have no direct experience with this.
4. In SAS' eyes, they are giving SAS away to Universities and educational
institutions. The costs they are recouping there are mainly documentation,
distribution and support costs. Even though most have central support units
within computer centers and such, supporting campuses of 35,000 students can
be a bit overwhelming. I have always been a proponent of giving it away at
the university level, training faculty and graduate students on the tools
and then reaping the benefits when they join the workforce. I'm not sure
that's how it always gets implemented though. I heard a recent story of a
prof wanting to use Enterprise Miner in a graduate course and SI would not
provide them with a copy. shame shame.
5. Third world countries... hmmm not sure what to say here, expect you're
probably right. I remember receiving a call from an agricultural institute
in S.A. about 5 years ago that asked if we could help them recompile this
program on their brand new high performance computer (by our standards, 1st
generation boat anchors). The software: a version of SAS that was
distributed to agricultural experiment stations in the late 1960's call the
Statistical Analysis System. Oh, but it had it's own manual. ;-)
6. I would think that supporting Version 7 or 8+ on Linux would be easier
than supporting older versions. Tech support would have to train folks on
Linux and because of turnover, also have to train them on 6.0x. A standarad
Linux would be nice (say Red Hat), but getting the Linux community to give
up their Slackware in favor of standardization would probably bring the
walls crumbling down.
7. The marketing possibilites for SAS internationally in third world
countires intruigues me. I can see the press release now:
WHO and SAS Solve World Hunger
Cary, NC - Today SAS Institute and the World Health Organization solved
world hunger by providing a steady revenue stream for third world countries.
This new program will allow street vendors and marketeers to seel bootlegged
copies of the SAS System to Linux users. Due to to the United States stance
on 128-bit encyrption outside the United States and world-wide preddominance
of the Linux operating system, industry analysts predict that the
re-invesment in local economies, will boost the third world GNP to twice
that of Japan and U.S.
I'm not sure that I've introduced anything new that Peter Crawford, Joe
Kelley, Karsten Self or Tim Churches hasn't thought of, I feel compelled to
join the Linux bandwagon. Afterall, a departmental sized server that is
low-cost, remote administration capabilities and has both command-line and
GUI interface does not exist. ;-)
Back to my NT box... for now.
--greg barnes nelson
Author of: Exploiting SAS in Server Environments: A UNIX and NT primer
(To be released after the Y2K thing has been fixed.)
-----Original Message-----
From: Tim Churches [mailto:tchur@BIGPOND.COM]
Sent: Sunday, March 07, 1999 2:41 AM
Subject: A version of SAS which runs under Linux?
There has been considerable discussion on this list about SI's apparent
intransigence (or is it diffidence or indifference?) on the subject of a
version of SAS for Linux, despite the rapidly growing support for the Linux
operating system. This growth in support includes not just individual users
but also many of the world's largest software companies (excepting THE
largest...), who are actively promoting Linux and are releasing native Linux
versions of their flagship products. Further discussion on SAS-L has
focussed on a suitable marketing model for SAS for Linux, given that a great
deal of software for Linux is free, including many of the Linux versions of
software released by major software companies.
One suggestion has been that SI release a very cheap or free version of SAS
for Linux which is based on an earlier release of SAS so as not to entice
potential customers away from paying for a more recent version.
Some of the benefits of such a version of SAS, which I will refer to as
"free SAS" might include:
a) widespread adoption of free SAS in educational institutions, where SAS
has been at a disadvantage due to cost issues (even with academic discounts
for SAS license)
b) widespread use of free SAS in developing countries (and in impecunious
NGOs in developed countries), where SAS has always been too expensive to
consider
c) the training benefits of a) and b), since shortage of SAS-literate
personnel has often been a barrier to the adoption of SAS in corporate
environments
d) the humanitarian benefits of b)
e) a firm presence in the Linux mindspace, given that Linux looks like it is
about to take-off in a big way
f) lots of opportunites for cross-promotion of commercial versions of SAS.
It occured to me that maybe SI has a version of SAS which would be just
right for this purpose: SAS Version 6.04 for MS-DOS.
Although I haven't tried it, SAS Version 6.04 for MS-DOS will almost
certainly run under Linux on Intel-based machines (which is the vast
majority of Linux machines) using the DOSemu subsystem (DOSemu is a DOS
emulator which allows MS-DOS programmes to run under Linux - see
http://www.dosemu.org for details).
All that SI would have to do is verify that SAS V6.04 works reliably under
DOSemu (volunteers may be able to help with this task) and then just
recompile it with the setinit licensing bits removed. Therefore the
development cost to SI would be minimal.
Of course, SAS V6.04 for MS-DOS also runs under MS-DOS. There are many more
286, 386 and 486 computers running MS-DOS (+/- Windows 3.x) than there are
Linux boxes. Old PCs running MS-DOS are the norm in many developing
countries as well as many educational institutions. Therefore, releasing a
free version of SAS V6.04 for MS-DOS would also acheive benefits a), b) c)
and d) above far better than a native Linux version of SAS, while still
making many Linux users happy.
Yes, I know the idea of having to use an MS-DOS program is anathema to many
Linuxistas, but the rest of us are not really interested in such computing
theological issues and just want to get the job done with the best balance
of cost, reliablity, speed and facilities.
Personally, I think that SAS V6.04 running under DOS emulation under Linux
would be just fine for very many purposes. However, after using Version 6.12
or V7, very few users would want to go back to using V6.04. That is the
point: free SAS V6.04 would be adequate for many uses where SAS has been
ruled out because of cost, but not very attractive to the majority of users
who can already afford more recent versions of SAS (or who need SAS on
heavyweight hardware like mainframes). Of course, some of the new SAS users
which free SAS would attract would ned up migrating to a commercial SAS
license.
For those who are familiar with SAS V6.04 for MS-DOS, it was one of the
first versions of SAS (actually 6.03 was) which was written entirely in C
and was a precursor to Versions 6.06 onwards on multi-user platforms (Unix,
VMS, mainframes) and Windows. SAS V6.04 shares almost all of the same data
step and procedure syntax as SAS V6.06 or later, but lacks things like
dataset indexing and compression. It does not have a true GUI interface -
the choices are a command line interface for batch mode or a clunky
character-mode interface with the usual "3 windows" (Program Editor, Log and
Output). It does include SAS/AF and SAS/FSP but does not support Frame
entries - the interface would be familiar to users of SAS on mainframe
terminals i.e. decidedly utilitarian. Production of high resolution graphics
is supported through SAS/GRAPH but there are no drivers for GIF or other
Web-related formats, although most importantly PostScript output is
available (which can then be converted to just about anything else thhrough
post-processing). Various modern statistical procedures are absent, such as
PROC LOGISTIC and PROC GENMOD. However, there are numerous free DOS-mode or
native Linux alternatives for these available.
So, how do people feel about this proposal? Should we petition Dr Jim?
Cheers,
Tim Churches
Sydney, Australia
[text/html]
