[Cadre-politics] policy decision: giving up hosting

[Dan MacNeil]

The CSL board is discussing ending our email/web hosting. They 
will probably make a final decision in March.

We want your comments, especially from customers. We'd even buy 
lunch for commentors who didn't want to take the time to type out 
their thoughts.

Some board members (Fred especially) have had some very useful 
and thoughtful things to say. Feel free to neglect your children 
or your day job to say them again or to point out flaws in my 
reasoning, assumptions or facts below.

WHY END?
In 2002, to get the services we provide, you'd have to pay $50
per month. These past couple years service equal to ours was
available for $10 per month. Now there is at least one group
(Dreamhost.com) that in terms of uptime, number of services and
timeliness of tech support, is better than we are and (for 501c3)
organizations no cost.

We don't make enough money from hosting related donations to even
pay for equipment. Generously, we'd need 15 times the voluntarily
paying customers we have now to pay for equipment and VISTA fees.

Hosting takes time from important stuff like improving 
http://mvhub.com, creating websites (which we don't do at the 
movement), fixing people's office computer problems or 
contributing to free software.

WHY NOT END

WE ARE BETTER IN SOME SMALL WAYS
There are a few quirky things we do better. For example we allow
people to have 6 Gigabyte email boxes. Sometimes we work
directly with register.com to renew people's expired domains, a
month before a big festival.

UP-SELLING
One place where we could become "sustainable" is LAN/Desktop
support. In a market where the going rate is $70/hr, we can
charge $35/hour and have money left over to do other stuff. The
most likely paid support customers are the people we're
currently providing service to for free.

Starting from scratch, Dreamhost does a better job than we do. 
However, the hassle of switching providers is unlikely to 
pre-dispose people to our sales pitch on desktop support.

TRAINING AND PRACTICAL EXPERIENCE
Five people told me that their sys admin experience  with us on 
hosting got them a market rate job. Not everyone comes to us with 
the skills to do software development. If we don't do hosting, 
the population of people who we can provide training and 
practical experience grows smaller.

Unlike the equipment in NPO offices, Hosting is something we 
control completely. We have separate testing and production 
environments. We have documented procedures and quality assurance 
  checklists and scripts.

Unlike a visit to somebody's office to fix some mysterious 
failure, the risk of somebody screwing something up in our 
environment is relatively small.

SOMETHING WE CAN TOUCH
It is easy to explain "We provide webhosting & email" to funders. 
Hosting is big, visible and easy to understand. Seeing it, doing 
it, touching it makes us all feel connected. Talking to somebody 
in a call center or (worse) working in a call center is not a 
connected feeling.

http:///Habitat.org builds houses inefficiently. The illusion is 
that mobs of unskilled volunteers build the houses. The reality 
is that a hiring a backhoe operator is more efficient than buying 
lunch for 20 people. For every 5 engaged construction volunteers, 
there is a skilled worker, pre-working the work. The efficient 
way to build things is to hire a contractor.

Of course if you have friends at a church willing to make lunch 
and don't have cash, doing things inefficiently is better than 
not doing them.

More important, Habitat is not especially concerned with building 
houses.

Habitat's goal is to eliminate poverty housing from the face of 
the earth. Habitat's sub goal is to make simple decent housing 
for everyone a matter of conscience.  People develop this 
conscience by putting their hands on boards, working with 
prospective homeowners and getting sore muscles.

Our goal is to create a society where people's status comes from 
the value of what they give and achieve instead of what they own.

Giving away free software and effective technology support is a 
menas to the end of a gift culture:

	http://catb.org/~esr/writings/cathedral-bazaar/

Parallels between us and Habitat may not work the way I claim. 
Frequently Habitat affilites sub-contract the tricky and lawsuit 
inducing work like electrical wiring. We may be able to 
accomplish our goals without our own chip fabrication plant.

WHO OWNS THE SOFTWARE SERVICE
For a few years now, the computing trend is software as service:

	http://paulgraham.com/road.html

The question is:

	Who owns the service?

It sure would be nice if 100 years ago, the push had been to 
household size electrical generation.

HOSTING ECONOMIES OF SCALE ARE NOT STRUCTURAL
Twenty years ago, to run a cell phone, you needed a computer the 
size of a small house.  Thirty years from now, 3 servers will 
have the CPU power of Google's 3 million CPU server farm. (see 
footnote)

In some industries, like wheat farming it is tough to compete 
without a few thousand acres of land and the money to rent a 10 
million dollar combine.

Small time operation that we are, we own about 15 servers right 
now. In 30 years we could probably afford 3 servers.

Given that thousands of people are working hard to create useful 
free software that we can use, google's edge over us is not as 
huge as it might appear.

The barriers to de-centralized, publicly owned software services 
can be scaled.

MY PERSONAL FEELINGS
I'm happy we have a board to make this decision.  Given our 
structure, it isn't my job to make this decision. It is my job to 
make this decision a difficult one.

If it were to me I'd flip a coin or make a decision by not making 
a decision.

"Pick one or pick none"...is a real concern. Given limited 
resources, better to do one thing well than several things half 
well.

If we don't host, we can focus on http://mvub.com . There are 
some alteratives to MVHub (211) that aren't nearly as nice. 
However they are close to good enough and they have a lot more 
money behind them. There is a danger they will supplant us 
outside the Merrimack Valley.

Ending hosting feels like closing a door.

I'm often the one who hangs on past the point of sanity.

FOOTNOTES

Moore's Law: CPU power doubles every 18 months.

30 years / 18 months = 20 doubling periods

CPU power will double 20 times in the next 30 years.

2^20 = 104,8576

3 million / 104,8576 is 2.86 servers