On Sun, Dec 30, 2012 at 12:28 PM, Troy Benjegerdes <..hidden..>
As a business owner (farming and technology consulting), I have the
1) educate business owners on the benefits of the Debian distribution
to allow long-term (30 year) support with no forced upgrades.
2) provide community support for 18 months
Probably be 36 months or so.
3) distribution support for the distribution's term
4) Provide offers for long-term support contracts (based on Debian)
so that I, as a business owner, have a full, up-front accounting of
the NPV of running whatever version I happen to be running on the
latest 'community' supported version/distribution combo.
I think one has to understand though what the cost of running an accounting system on a computer for 30 years is.
As background I want to say I have a tendency to run computers for insanely long times. My firewall ran a Pentium 1 chip until the chip and hard drive both gave out in 2006 several years after the fan stopped running. The computer would have been at least 12 years old and the approach to firewalling that I took with that system is similar to what I would take in other areas today.
Similarly I have been running Linux servers since 1999, and have some ideas of how security concerns have changed in that time.
With our current codebase, the basic problems you will run into are as follows:
1: If you don't upgrade components, you set yourself at risk for security issues.
2: If you do upgrade your software components but not your hardware components, eventually you may find that the hardware just isn't sufficient.
3: If you upgrade your hardware components but not your software components, eventually you may find that the software won't run on the newer hardware and then you will be forced to upgrade.
I don't doubt that in some cases, a well engineered server might run with minimal interruption for 15 to 20 years, but beyond that I would begin to doubt that.
As some background: I can buy a 30 year old combine for around $5,000,
and then fix it up, and the quality of the grain it harvests is exactly
the same as a brand-new $250,000 John Deere.
This is true. But they operate in very different environments. I doubt any server hardware of today will get 30 years of life and so the complex interactions between software and hardware have the potential to be somewhat problematic.
In 30 years I expect to be fabricating silicon for replacement CAN
controllers because at some point I'll be able to pick up some combines
that are fully functional, except for lack of a chip that nobody makes
I would really like to see a 30 year NPV calculation of the total cost
of ownership of an up-front investment of say $10,000 in ledger-smb
consulting and setup, and running it for 30 years, vs what the NPV of
the upgrade treadmill of windows/macos/cloud hosting/quickbooks/whatever.
It would be interesting to look at this in terms of both upgrading and non-upgrading. Obviously upgrading too often increases the total cost of ownership, but not upgrading often enough may do the same.
I think the formal 'end of life' proposal is that there *is* no end
of life, so long as you are happy with how everything works, and
understand you need to run this system on an isolated network that is
firewalled by something that *does* get security upgrades.
This is true, and is largely the point made above.
FWIW, I don't envision turning anyone away from my own consulting services for running software that is too old. The thing is that at a certain point there are additional costs that get incurred and as a consultant I would be remiss if I did not advise my clients that an upgrade might save them money and hassle in the long run.