[wplug-plan] September speaker

Bryan J. Smith b.j.smith at ieee.org
Wed Aug 29 07:02:20 EDT 2007

First off, you'll find when I approach something, I like to be
all-encompassing.  I.e., I typically have a 3-prong approach ...

1.  Market-deprogramming

I break down "products" into "technologies" and "solutions."  It's
essential to understand the full spectrum of technologies and solutions.
That's the only way you know if a product will work or not work for your
organization.  Especially in the "open systems" world where you often
use multiple services/solutions.

2.  Enterprise best common practices (BCP)

The essential do's and dont's in an enterprise.  I still find way too
many enterprises don't know what they are doing.  E.g., for disaster
recovery, the question isn't "Is tape good or bad?  Should I be using
disk instead?"  The question is, "Am I addressing on-line, near-line and
off-line recovery -- all 3 -- correctly, and if not, how should I?"
You'd be surprised how _inexpensive_ it is to "do it right" and it's
_cheaper_ than how most do it when they are not addressing near-line
and/or off-line correctly.

3.  The SOHO cheat sheet to enterprise on a budget

How to implement sound solutions without spending much.  Many SOHO
solutions cost way too much and do too little.  There are sound SOHO
solutions and practices that people don't even think of.  E.g., with
regards to storage, I've also seen departments that use USB and other
storage and then wonder why their servers are not reliable and crash
regularly.  Simple practices.

On Tue, 2007-08-28 at 17:22 -0400, Teodorski, Chris wrote:
> I'd love to hear something on SOHO disaster recovery strategies...

As above.

On Tue, 2007-08-28 at 14:00 -0400, curlynoodle at gmail.com wrote:
> I like the idea of "Small-footprint x86 hardware considerations for Linux 
> appliances", but enterprise authentication would be a good topic too.

There's a lot out there in x86 SBC and small form-factors.

E.g., the AMD Geode NX gives you 9-issue Athlon power for little cost
and power.  And then there are the network, storage, boot and other
considerations.  You shouldn't expect a commodity x86 Linux appliance to
rival anything with Network Processing Elements/Engines (NPEs), but they
can do a lot, in small form.

The nice thing of staying with x86 is that you don't have to (although
sometimes you still want to) build a separate toolchain, as well as the
fact that they target can be self-hosting / self-developing.

On Tue, 2007-08-28 at 13:54 -0400, Patrick Wagstrom wrote:
> Thanks for responding.  My vote would be for the second topic on
> enterprise authentication.  Do other folks have opinions?

That would also have to include network object naming and store
(directory) as well.  Without network object naming and stores,
enterprise authentication becomes impossible -- or should I say --
enterprise authentication becomes rather limited and not encompassing
for your enterprise.  I've seen that too may times.

Also understand this is short notice, so I can't show much "hands on,"
but would be more "best common practices" (BCP).  Especially since I've
had to re-load my notebook with Fedora 7 32-bit, whereas I had Fedora 7
x86-64 and all sorts of stuff running like Fedora Directory and
Certificate Server (long story, I couldn't get some of the tools at my
new job running under x86-64, not even after extensive 32-bit library

I'll break down Microsoft ActiveDirectory Services (ADS), Novell
eDirectory and Sun One into their base components, and show the open
technologies that are their make-up.  I'll then dive into scenarios.
People also don't know that the _entire_ and _original_ "enterprise"
LDAP and Certificate system, the iPlanet Directory and Certificate
services, has been GPL/MPL'd since 2005 by Red Hat.  I.e., Red Hat
bought it from AOL-Netscape and it's far more "experienced" (since the
'90s) than OpenLDAP at these scenarios.

If there is one thing I can't stand in enterprises, it's those who
follow a "cookbook" method of making Linux into ADS' "bitch."  The Samba
suite of services is _not_ required to work with ADS, and the suite
itself is rather "limited."  Most real enterprises have "peer" systems
that replicate between each other, one "open" and then one
"proprietary/hostageware."  The reason is that the latter often requires
massive infrastructure changes every 2-5 years, the former lasts decades
(as most large enterprises and universities can attest).

Bryan J. Smith         Professional, Technical Annoyance
mailto:b.j.smith at ieee.org   http://thebs413.blogspot.com
        Fission Power:  An Inconvenient Solution

More information about the wplug-plan mailing list