[WBEL-devel] Long- Thoughts on WB maintenance

[RftD]

On Thu, 27 Nov 2003, Charles Lacour wrote:
> $350 per box is very workable for 30-40 servers for the entire company. $350 
> per box (per year!) for 2000 desktops is WAY too expensive.
> 
> WeŽre not going to pay that $350 no matter what. Maybe we wonŽt switch to 
> Linux on the desktop, maybe weŽll go with some other version, maybe weŽll go 
> with WBEL. Regardless, Red HatŽs not out any money, because we were never 
> going to make that purchase to begin with.
> 
> I think Red Hat got a little greedy ($350 for RHN only is pretty steep), but 
> the main thing they did is say that "we want out" of the low-cost, low-income 
> part of the market. (RHN-only accounts, and no-support-at-all installations, 
> basically.)

I think you're quite right there (except perhaps the greedy thing), in
that desktops are a real issue for a lot of people.  A company can go out
and buy desktops (monitor, MS windows software, etc) for $800 each, and
keep those for at least 3 years.  But then the software cost would be over
$500, (lets stick with the WS3 product on desktops rather than ES) which
seems pretty steep compared to just sticking with the MS product that 
comes with the box (and most major manufacturers come with MS on all 
desktops).  

> I think they decided to concentrate on the high-margin stuff, where they could 
> easily justify hiring another engineer any time they needed one, because each 
> engineer was costing them $75,000 (to pick a number out of thin air), but 
> bringing in $400,000 in revenue. 
> 
> If they're dealing with lots of "little people", the ones to whom $350 a year 
> is a big deal, they're spending $75,000 for maybe $80,000 worth of revenue. 
> (And maybe less than $75,000, which means the more they "expand", the faster 
> they go broke.)
> 
> My point with all this is that WBEL (and any similar product) is not a threat 
> to them. It's a portion of the market they decided they wanted to get out of. 
> If WBEL is wildly successful, and Red Hat sees an opportunity there, all they 
> have to do is start allowing people to install it wherever they want (but 
> still having to pay to get support), and WBEL is out of "business".

I think you are quite right about this.  The hosting company where I work
took a similar position, after aquiring the hardware/software from a
smaller hosting company that was a client of ours when they went bankrupt.  
The product obviously doesn't make much money (in fact, the revenues are
about 1/20th or less of the costs to keep it active) but we were able to
shift the really really small customers off to this product, with the
understanding that they get *no* tech support in exchange for the absurdly
low price.  Even though we're losing some money on the product, we'd
rather than them use that than have to support them.  The support costs
for these clients were several times more than we're losing on the minimal
maintanence we do on the box.  They would often feel (when paying our
normal prices) they deserved to have us spend hours on the phone
explaining how to connect to email using outlook.  Since our primary
expenditure is personel, this was becoming a problem.  We did a cost
analysis and determined that 5% of our revenue was costing 50% of our
support costs.  Now we've lost probably 4% of revenue, but the costs
savings more than make up for that.

> It's my opinion that that is exactly what they SHOULD have done. Have a more 
> reasonable cost for RHN, no cost at all if you are willing to do all the work 
> of figuring out what to upgrade when, and then the higher-priced stuff for 
> those who really need the higher-level support.
> 
> What I'm really worried about is that people will go to something like SUSE 
> for the low-end stuff, and then decide that sense they're running SUSE on the 
> low-end, they should go ahead and run it on the high-end, too.

Well, I'm not sure Suse will be the primary concern, though they are $99 
vs $179 for Suse Desktop vs RH WS.  More likely dangerous is that Suse 
Personal is only $39, which means at home I'd be more likely to use that, 
and then when I go to work I'll feel like recommending suse because I use 
it at home (or school, etc).  Another big mistake (IMO) is that RH WS for 
AMD64 is $800 compared to $120 for Suse.  That means when I get my AMD64 
based desktop (which are available in the $1000-1400 range for a 
reasonable system) there's no WAY I'd install RH on it.  If AMD64 proves 
to be successful, hopefully RH will come to its senses, though this is 
difficult once you've already sold the product for 8x as much for a while.  
Although Fedora isn't a bad idea, especially for people coming from 
windows platforms, a 3-4 month upgrade cycle is way shorter than I'm 
willing to put up with, and they don't have an AMD64 release yet anyway.

> Two or three years down the road, Red Hat could be seeing their enterprise 
> business going away, simply because they can't cover the full range of 
> people's needs.
> 
> Whitebox Linux would actually HELP that, because then companies could say to 
> themselves "it's really the same thing -- what my people know on their 
> desktops will work just the same on the server", and they'll happily keep 
> using WBEL for free on desktops, and similar "unimportant" boxes, and buying 
> RHEL for their important machines.

And the technical people will also be more likely to try it out at home
and be familiar with it.  I can't really see recommending Fedora to anyone
other than developers, given the unreasonably fast upgrade cycle.  I know
many people still running Win98 at home, so I doubt they'd want to upgrade
to Linux and get the "feature" of having to upgrade several times a year.

> John (and everybody else who's contributing) is doing this precisely because 
> Red Hat doesn't want to. Most of us would be much happier if this had never 
> been necessary, and if the existence of WBEL rubs Red Hat's nose in the fact 
> that they screwed up in this area, wonderful.

I know that's how I feel as well.  I like RH.  I use it on all our Linux 
servers now, and on every desktop I can.  But for the system that acts as 
a vpn server from my home to our office, the cost just isn't justified.  
The system is running on ancient leftover hardware, and buying RHEL would 
probably cost more than the hardware is worth.  I don't want to use Fedora 
because I'd like to make sure the box has patches available (even if I 
have to compile them myself) so at the moment the two systems are running 
RH9 with RHEL3 RPMs to update them.  This is a fine short term solution, 
but as time goes on it will be harder to keep things working.  Having WBEL 
is a great solution.  Otherwise, I'd be using debian or whatever, and 
wondering why I have to learn something new.

> The idea that Red Hat is going to get protective of "their" code is extremely 
> unlikely. They've demonstrated time and time again that they understand 
> open-source, the GPL, etc.
> 
> It's probably not possible for them to put out a version of Linux that 
> couldn't be "whiteboxed". Even if it were possible, it would be suicide, 
> because the open source community would shun them for violating the rules. 
> Since the open-source community does 85-90% of the work in a Red Hat 
> distribution, Red Hat would incur hugely increased costs if they tried it. 

The real problem I see with redhat releasing their own product is that it
would dilute their trademark/branding.  I assume that is *why* they
switched from regular RHL to calling it Fedora.  They want to identify
themselves with enterprise class software.  If they release other things,
especially if they are much more popular, that's what people will
associate with them.  So they just drop the non-enterprise products.  
They may even be willing to "covertly" help us out a little, if the
purposes of the drop were branding and support costs, by doing things like
separating out their trademarked images etc into an extra source file,
though this would be a little extra work for them.  It all depends on how
they feel about WBEL and how they feel about their own product.