Subject: Re: a tool free software developers need
From: Keith Bostic <bostic@sleepycat.com>
Date: Mon, 18 Feb 2002 10:24:02 -0500 (EST)

>>> I suspect the most expensive part of the operation will be to pull
>>> together sysadmins with the wide range of talent you'd need.
>> 
>> I'll bet you're right.
>
> I'd guess having a wide range of different hardware would also increase
> costs, e.g. spare part stocks, compared to a rack of identical servers.

Sys admins aren't the cost.

I've kicked this idea around with two different discussion groups, and
in both cases we came to the conclusion the way to make this work is a
single disk you re-install for each new user.  So, you get the RedHat
release N CDs, install on a single system, and from then on when someone
wants that release, do a disk-to-disk copy of the vanilla installation
onto a test system's disk and boot it.

No large system admin cost, because users of this kind of setup want as
vanilla a system as possible (since vanilla is what they'll see in the
field).  So, a sys admin who installs new releases on each machine type,
and never touches them again.

That said, neither group made progress.  The initial hardware/bandwith
costs, although not enormous, are real, and nobody could be convinced
there was money here.  Open Source groups (a) rarely test their software
and (b) don't have any money to spend on testing their software, so it's
small software companies that might buy the service.  Since I can buy
a random x86 and do the same install myself for about $800, it's going
to be hard to get people to pay a lot of money for the service.  There's
some added value since purchasers of the service could easily test on
lots of minor variants (RedHat 7.1, 7.2, 7.3 and 7.3a, or, gcc 2.95.2,
2.95.3, and 2.95.3 plus patch 37), but again, it's not something people
will pay a lot for -- most companies test RedHat 7.1 and assume, with
good probability, that any application that runs on 7.1 will run on 7.X.

Anyway, the only hope I have for this kind of service is that someone
will do it for themselves and decide to let other people use it.  Not
because they'll make money, but because it's a useful thing to give to
the community.

Regards,
--keith

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Keith Bostic
Sleepycat Software Inc.		bostic@sleepycat.com
118 Tower Rd.			+1-781-259-3139
Lincoln, MA 01773		http://www.sleepycat.com