[Foundation-l] Re: Our exponentially increasing costs

[Anthony DiPierro]

On 10/23/05, Daniel Mayer <maveric149 at yahoo.com> wrote:
>
> --- Dori <slowpoke at gmail.com> wrote:
> > Yes, but when you enter into a legal contract, it's not just you that
> > we have to worry about, it's the other party as well. I really don't
> > think we'll be getting much revenue from this deal to justify the
> > risk, and in general I don't think in the *long term* relying on
> > advertising will keep Wikipedia afloat.
>
> Er. Not that we'd ever do this unless *absolutely* necessary to keep us
> online,
> but do you have *any* idea how much advertising revenue would be brought
> in by
> an Internet top 50 website? At least a couple hundred thousand dollars a
> month
> for something relatively unobtrusive like Google Ads.


It boggles my mind that this isn't being done. Seems to me like a totally
irrational fear that Wikimedia would somehow start catering to the
advertisers. It's almost as though Wikipedians are afraid of money or
something.

> I'd like to see us plateau on the Internet first, and then worry about
> > expanding off-line.
>
> We have been waiting for this mystical plateau for years now. The slowest
> I
> imagine we will ever grow is to just keep pace with the Internet's growth,
> which, btw, is also exponential but not nearly as strongly exponential as
> our
> current growth. So we would need to wait for the whole Internet to plateau
> in
> growth, which is still decades in the future. We can't wait for that.
>
> > Off-line distributions will be a lot more
> > difficult and a lot less useful. On the other hand, there is no reason
> > why we couldn't start on Wikipedia 1.0, and I've yet to see that get
> > off the ground.
>
> There is no reason why we can't do both in parallel. One will benefit the
> other.
>
> > I can't believe that we would need to keep growing at 1-2 million
> > dollar hardware acquisitions every year. You can't go much higher at
> > some point.
>
> Er - we will need to spend at least $2M next year and a lot more the year
> after
> that if our growth curve holds. See my email about our *exponentially*
> growing
> costs.


I seriously doubt there's a *need* to spend $2 million next year. At the
very least, you could be leasing servers rather than buying them. Buying
servers doesn't really make that much sense at this point in the growth
curve, unless you're going to take out some loans. You could even look into
outsourcing the whole thing, because one employee simply isn't enough to run
a multi-million dollar server farm. There are people who can do this more
efficiently than you. Hire them or outsource the job to someone who will.

In the longer term, there's always the ability to go with a more distributed
system, with volunteers hosting the content via a P2P type network. Not pure
P2P, of course, more like a BitTorrent type system. In the past when this
was brought up the answer given was that it wasn't necessary, originally the
answer was "Jimbo's paying for the bandwidth anyway". Well, now it sounds
like it's getting to the point of necessity.

Ultimately I think there are only two real solutions: build a truly
distributed network or get some sort of sponsorship (or as the cynics like
to call it, advertising). The other possibility, grant money and donations,
probably isn't going to come in big enough numbers to run such a big website
without some restrictions on its use. And this is a good thing. We shouldn't
be pouring money from donations into running a website just because we're
too scared or proud or whatever to put a little text acknowledgement of our
sponsors at the bottom of our pages. Grant money and donations should go
toward creating new things, not maintaining the status quo.

> It would help to keep features from creeping in that
> > require more hardware too. I personally never liked the big use of
> > templates and categories. It's become a lot harder to make sense of
> > what's going on in an articles that use 10 internested templates (some
> > with if-else logic no less), and there is no need for categories when
> > you can just link (and besides I hate the idea of pigeon-holeing an
> > article into a category to begin with).
>
> Dramatic increases in readership, not feature creep, is what is driving
> hardware cost increases. Again, see my other email.
>
> -- mav


As someone else said, surely the features aren't helping. But with good
design, you can have features and efficiency. I'll admit that there has been
some progress in this regard in the past year, but there are still some
major design flaws in the mediawiki software and the overall architecture of
the system which are causing the hardware costs to be much more than they
should be. This might have been reasonable in the past when the cost of
hardware was low, but those days are over. At least there's a CTO now,
though his user page curiously says he's "On WikiBreak". And considering
that I don't recall there ever being a period where Wikimedia advertised
that it was hiring this position I have my doubts that they can be so sure
he's the most qualified person for the position.

Anthony