[License-discuss] code hosting (was Re: Evolving the License Review process for OSI)

[Rick Moen]

Quoting John Cowan (cowan at ccil.org):

> A basically volunteer agency like OSI "self-hosts" when someone volunteers
> to host for them.  When that volunteer loses interest, the "self-hosting"
> goes away.

Failover is a thing, isn't it?

Let me tell you a story.  I'm a longtime volunteer with a local
all-volunteer-owned-and-staffed annual convention for (mostly literary) 
science fiction -- that shall go nameless here -- in my Silicon Valley /
suburban-San Francisco area.  The convention had followed (bad) advice
to move all of its Internet presence to a bottom-dollar
WordPress-specialty hosting company, Bluehost, but found that (as I
could have predicted) the firm was abysmal at everything but WordPress,
and in particular seemed unable or unwilling to run a reputable SMTP
operation that avoided being shunned on DNS blocklists as a spamhaus,
where the Webmin-based customer WebUI required for all customer
administration of hosted domains broke frequently, and where customers
had no access to logfiles for solving problems.

Anticipating the convention's 'What would you suggest we do, instead?'
question, I built a prototype Internet server using Debian in a VM
container, that did full smarthost SMTP with good spam rejection,
Web-based administration of mailing lists, and MoinMoin (a simple
Python-based wiki) for ancillary Web content.  And, critically, I built
in failover.  And service monitoring and some degree of update-checking.

The predicted question arose, and I said 'You could run _this_ [showing 
how it worked] installed on a junk 2010 box with half a gig of RAM from
Weird Stuff Warehouse, running on a static IP in someone's garage.  And
you would have a second instance running on another junk 2010 box on
static IP in a different person's garage.  The first keeps the second
updated, and if either fails, the other takes over, a concept called
"failover".  And then if necessary you find a third garage for a
replacement failover.'

They didn't like it.  'Whom would we get to run this?'  'Mostly nobody.  
It runs; you avoid fooling with it.  Once a month of so, you find
someone with enough clue to type 'apt-get update && apt-get dist-upgrade'
as the root user.  Otherwise:  Just.  Don't.  Fool.  With.  It.'

'But I don't understand.  Don't we need an expert to run it?'  'Not
really.  Just don't fool with it.  If you need to, you're it freakin
Silicon Valley; this really isn't difficult.'

A suitable, boringly reliable soution didn't suit them.  They wanted
Someone To Sue (http://linuxmafia.com/~rick/lexicon.html#someone-to-sue).
Last I checked, they were still suffering with Bluehost because it costs
money therefore _must_ be good, and occcasionally outsource to Mailchip
when the cognitive dissonance about poor Bluehost SMTP deliverability 
becomes too big a problem.

And yet, failover actually _is_ a thing.