[nSLUG] Introduction, Meeting? and Keane

George White aa056 at chebucto.ns.ca
Sat Aug 28 09:09:37 ADT 2004

On Fri, 27 Aug 2004, Jeff Warnica wrote:

> On Fri, 2004-27-08 at 09:04 -0300, Wayne wrote:
> > failing to run X properly with a garden variety Intel video 
> > card), the concept of an OS with everything installed and tested 
> > for the corporate environment (and then, why not home?) has to
> > enjoy some kind of success in the future.

Corporate environments are becoming too complex for any one-size fits all.
PC's are being configured with Human Resources and Finance applets to 
deal with vacation and leave processing as well as business function apps.
Conflicts with these (which are often purchased from a 3rd party) are a
big reason large organizations have to test patches and coordinate
security fixes with updates to corporate apps.  

Many universities are now doing their own linux distro (Knoppix is a
popular starting point).  I hope we will see more of that approach --
large organizations maintaining their own linux distro.  What you may
start to see is vendors of those HR and Finance systems doing linux
distros for customers that aren't large enough to have an IT group.
Those HR&Finance vendors are being burned by Microsoft because their
apps don't work with XPSP2 security features enabled.  

> Ah Sun. In this months Linux Magazine the opening editorial gives
> something of an overview of Suns seemingly random behaviour over the
> recent years, both WRT Linux and in general. I agree completely with the
> editorial.
> Historically Sun has produced some very cool stuff. On the high end,
> they have some very cool hardware. NFS is proof that they have been the
> driving force for some "open" systems. OTOH, why am I not using a
> graphical user interface with Display Postscript?

DPS has an excellent underlying graphics model, but allowing apps to send
arbitrary PS files to your DPS interpreter makes it too easy for an app to
grab excessive memory and or CPU resources.  Sun should have realized
this, but they were in a university/research mindset where they expect
users to change behavior to accomodate the system.  The solution is to use
flattened PS, which is one of the things PDF provides.  Adobe tried to get
unix vendors to replace DPS with a form of display PDF, but the timing was
bad and the price so high that Apple decided they could do it inhouse for
less (Apple already had the hardest part -- font rendering -- under
> Whenever a discussion about Sun comes up someone always mentions what I
> think of as their "primary feature", which I also think (can be) very
> important. That is compatibility. You can run exactly the same code on
> your $3000 Sun workstation and the $3,000,000 SunFire you have doing...
> whatever it is you do with a $3m computer. And you can do the same
> across time, you have backwards compatibility to as far back as when
> Sparc came out. Solaris/SunOS, perhaps even further back then that.
> Slackware not only no longer uses it, but doesn't even have the crypt()
> function. Still the default password hashing method on Solaris.
> What has Sun done in the past few years? They have introduced ia32/64
> servers, and a Linux based desktop. Hmm; that compatibility thing is
> blown to hell, no?

Java.  There was a recent article about Staples merging their web and
inventory systems.  One was using .net technology and the other Java
technology.  The .net experience was not good, so they moved everything
to Java based technology.
> There are other companies producing a (linux) OS with everything
> installed and tested for the corporate environment, and have been doing
> so for a while. Redhat and the independent SuSE. Independent Ximian too,
> desktop wise, anyway. What will be _very_ interesting to see is the next
> (well, first) major release of a fully integrated Novell/SuSE/Ximian.
> While, say IBM, has contributed enormously to the Linux/OSS community, I
> think it will be Novell who will get Linux onto corporate desktops in a
> big way.

SGI is doing interesting things with linux for really big systems.  They
are planning to introduce large (aimed at the >256 processor market)
parallel systems with vector and PAL coprocessing capability.  Ardent 
tried the vector thing 15 years ago but at the time there weren't enough
people who understood how to write software for such architectures.  SGI
has a slim chance of making this work, and when they fail some very
interesting boxes will be available cheap.

Right now, HP is the company that seems to be getting things together.
They moved ahead of Dell in corporate customer satisfaction and have
enough unix expertise to recognize XP for the abomination it really is.
They have Dec's compiler group which knows how to deal with the long
pipelines in the current P4's (at work, one of our models runs 2 processes
on a 3.2 Ghz P4 in the same time it took to run one on a 2Ghz pre-HT
P4 -- very impressive until you realize the old machine was a lowest bid
clone model with slow memory and the new machine was made by HP).
> Sigh. I dont know where Im going with this. But it has been a long, no-
> thinking-required, week. Call it a mental explosion.

Good time to take stock and make plans.  I'm thinking about doing more
work in APL (using the open source aplus).   I've spent a lot of effort
in the past decade trying to squeeze problems into hardware that was
just big enough.  Real soon now there will be working affordable hardware
10x more capable than the 5-year old machines I'm using now.  With the old
machines we had a limited toolset but now we are thinking about new kinds
of tools and the priorities are shifting to rapid development and away
from meeting hardware constraints.  

IT is moving to a new level of complexity due to the combination of
security issues and the need for scalability.  The companies that are in a
position to deal with this environment are IBM, HP, and Sun (interesting
that all these have strong compiler groups).  Dell, Novell, M$, etc. are
floundering.  Having a compiler group is a huge benefit because those guys
have to make the hardware perform -- both on SPEC and customer code --
so are able to tell the hardware people where they screwed up but must
also deal with the OS, so understand where the OS gets in the way.  

George White <aa056 at chebucto.ns.ca> <gnw3 at acm.org>
189 Parklea Dr., Head of St. Margarets Bay, Nova Scotia  B3Z 2G6


More information about the nSLUG mailing list