on Fri, Jul 23, 2004 at 07:40:58PM -0500, Kent West ([EMAIL PROTECTED]) wrote:
> Erik Steffl wrote:
>
> >Kent West wrote:
> >
> >>The result is that the vendor chooses not to go down that path, and
> >>Linux remains a niche product rather
> >
> >
> > let's see another example.
> Karsten made the very interesting point in one of his posts that the
> entire Oracle enterprise database software was ported to Linux with a
> single make command (although Paul indicates that the reference is not
> easily found, and I haven't bothered searching for it).
I'm stuck on a system that makes cut'n'paste difficult ATM, and my word
stemming wasn't fully accurate. "just typed make" (quoted) plus "Linux"
should turn up some interesting reading though, and not just Oracle.
> What this indicates to me is that vendors could write for all the
> nix's rather trivially, if they would just bother to do so. If that's
> the case, then yeah, the claim that there is no single Linux is a red
> herring.
There's a couple of ways to look at this.
Cross-platform development may not be entirely trivial, particularly as
an after-the-fact add-on, but:
- If you're working with a codebase that *has* been aimed at
cross-platform targets for years...
- If you're adding a new platform that is largely similar to the bulk
of your existing patform targets...
- If you've focussed on modularizing your code, isolating exceptions,
and writing to standard rather than "special" (aka undocumented,
aka vendor-specific) interfaces and APIs...
...then the task of adding a new build target is pretty low.
Contrast that with Microsoft's *own* experience building to seperate
codebases. My understanding is that MSIE for the Mac is an almost
wholly different animal than the legacy MS Windows version, and suspect
that this holds true as well for the Solaris port (since abandoned), as
well as MS Office ports to Mac.
Companies which have long supported both Windows and Unix products
generally found Linux support pretty trivial (though in the consumer /
end-user space, finding paying customers could be a problem -- there's
so much good stuff out there). Those which had dropped their Unix
support, or never had it in the first place, were caught flat footed.
And so, Microsoft with it's 93% niche of the desktop market (and
substantial, but smaller, 30ish% niche of servers) will be a shrinking
ghetto for these outfits. Should be interesting to watch that play out.
But then, Linux and FSF free software have been changing the software
market dynamic radically over the past five years, and will only
continue to do so.
> My impression (as a non-developer) for several years has been that the
> differences between various distros of Linux is a problem for vendors of
> non-open software. Perhaps then, the problem is not that there's no
> single version of Linux for vendors to standardize on, but rather that
> there's a _perception_ among vendors that there's no single version of
> Linux.
What's hard for the techies to realize is that this becomes a
partitioning problem several different ways.
Overall, in server space, Linux is ~30%.
That's partitioned into Red Hat, SuSE, Debian, and pretty much
everything else.
So now you've[1] got to:
- Target sales and marketing to some subset of the 30% Linux share.
- Develop any specific code required for each subset of marketshare.
- Create support systems and infrastructure for each subset of
marketshare.
- Develop co-branding programs, partnerships, and other corporate
relationships for each subset of marketshare. For companies, it's
often these relationships that drive deals.
Given that *everyone* is working on limited resources, *and* that the
corporate environment is simply bog-fixated on company-specific
branding, you're going to end up with "Linux is Red Hat", possibly with
a few saying "...and Novell/SuSE").
You and I know that the (technically) sane approach is to build to LSB,
offer RPMs and Debs (or offer source tarballs and let each distro build
what it wants), and increase your potential market from some fraction of
30% to a _full_ 30%.
However, if you approach Novell and Red Hat's marketing directors and
tell them "We've got an LSB-standard Linux product we'd like you to
carry", the response is going to be, "We'd prefer you tested and labled
this as 'Red Hat supported' (or SuSE supported, etc.), and request this
as part of our preferred partnership program". Not that I know the
specifics of language or programs, but it's going to go something like
that.
So for a lot of ISVs, it _does_ make sense, at the business level, if
not technical, to focus on one or two top branded Linux distributors.
Me? I think it's a long-term losing proposition. But for next
quarter's numbers, it makes sense, and that's where sales and marketing
folks, as well as CEOs and CFOs, are judged. So we'll see it for a
while.
> Part of this perception stems from earlier attempts with installing
> VMWare on a Debian box. The installer had to compile kernel modules,
> which involved me getting the headers (nope, that wasn't enough; get
> the full source), etc, and even then, the version of VMWare I last
> tinkered with wouldn't work with the newer kernel (2.6 maybe? it's
> been a while) that I had recently installed. So in that experience,
> VMWare wouldn't work on my particular "version" of Linux.
VMWare, along with Intel's compiler, is doing some pretty low-level
tweaking with kernel and HW-specific parameters. Alternatives such as
bochs and xen, while still emerging, illustrate better the free
software approach.
> 1) Could the vendor have made the installation more bullet-proof?
>
> 2) Without lots of resources to do so?
See above.
I think it's really marketing, as traditionally applied, that's the
killer here. And that doesn't mean that the marketers are wrong.
It *does* mean that to address the problem there's got to be an
alternative approach to marketing proprietary products on Linux, and to
selling specific Linux solutions.
I read Doc Searls with some interest, not becauls he's always right, but
because he *is* a marketing guy, with a long exposure and deep
familiarity with the business. It's not that he's always right -- I see
a lot of wishful thinking in what he says -- but he _does_ see many of
the failures of modenr marketing, and is cataloging many of the
instances of growing resistance to ever more ubiquitous and intrusive
marketing efforts. I find that it's interesting to note tht top folks
at LJ: Don Marti, a geek with very stong social justice stirrings, Doc
searls, the populists marketer, and Phil Hughes, who I can't quite put
my finger on, but as the guy who's relocated from Seattle to Costa Rica
in advance of the Great American Economic Crash may well be on to
something. If only that there are places wehere you can enjoy life
better.
> If the answer to these two questions is "Yes", then the original claim
> of not having a single Linux is probably just an excuse for lazy
> vendors. If the answer to either question is "No", then the original
> poster has a valid claim. I don't know the answer to these questions,
> but it'd be interesting to hear from developers who would know.
It need not be laziness.
The market-economics system is a *very* good way of optimizing for
specific goals and/or reward incentives.
Its very success at this means that it's also highly susceptible to what
might at first appear to be minor tweaks or perversions (aka
corruptions) of these incentives. You've got to be really careful what
you ask for, because you're going to get it in spades.
This isn't a criticism of a largely free-market based system. It *is* a
call to look at what guides the market, and to make sure that what
you're promoting is the Right Thing.
As I've noted above, a number of factors, mostly revolving around how
products are marketed, lead to a clumping or agglomerative affect around
a small number of Linux distros, largely commercially based. *If*
Debian sees as one of its goals the encouragement of third-party,
non-DFSG-free software availability, identified as official vendor
support, then we've got to look at what ways we can promote this, within
the existing marketing mentality and infrastructure. And for this, it
really *does* pay to understand how the other side thinks and is
motivated. Calling the behavior "stupid" is really a disservice to
ourselves. *Understand* the stupdity. Embrace it, and extend it ;-)
There's a lot of things about Linux which make it difficult for
marketing types to get their heads around it: numbers are notoriusly
hard to come by, and are notoriously soft. I tried pinning down that
"30% server marketshare" figure above, and with some googling, mostly
came up with results from 1998-1999 (height of the MSFT antitrust
activities and Linux dot-com hype) when everyone actually cared about
such stuff. One article noted that numbers are hard to pin down, and
likely undercount Linux (and overcount Microsoft) for numerous reasons.
As a guy who's spent many billable hours coming up with numbers for
marketing types, I can assure you that this is the sort of situation
which absolutely drives them nuts (and yes, I do take a certain perverse
pleasure in this knowledge ;-)
I had a conversation a few years ago with the founder of the
then-dot-com company I was then working with, about the merits of Debian
over Red Hat. "It's techncially superior" was my main argument.
"People don't care about technical superiority" was his rejoinder. "But
that's the whole advantage of Linux over Microsoft or proprietary Unix",
I said.
Short term, I've got to say that he was right. Not that it saved that
company. But Debian has shown persistant growth over the years, and is
generally the #2 or #3 distro in various surveys. And if you look at
the structure of responses, you'll find that it's highly represented
where quality and costs matter, *despite* having effectively no
marketing support or budget: technical shops, R&D, bootable distros,
and a number of razor-thin margin operations (unfortunately, several of
which have razor-thin moral martgns as well). But this is highly
reflective of the Linux space as a whole, say, in 1995/96, when I first
encountered Linux. Talking with a Stanford CompSci PhD candidate, I
heard that Pentiums running Linux were the preferred platform. This in
the then newly-christened Gates building. I figured I should probably
pay attention to what the Stanford CS grad department was focussed on,
and never looked back.
I think we'll win. I think the final victory will happen more suddenlly
than expected. I think that it will take longer to get there than we'd
like. But I'm OK with that.
Peace.
--------------------
Notes:
1. Assuming you're a proprietary ISV.
--
Karsten M. Self <[EMAIL PROTECTED]> http://linuxmafia.com/~karsten
Ceterum censeo, Caldera delenda est.
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
