Re: (eternity) Eternity as a secure filesystem/backup medium (fwd)

Forwarded message:

> Date: Sun, 18 Jan 1998 22:00:24 -0800
> From: Bill Stewart <bill.stewart@pobox.com>
> Subject: Re: (eternity) Eternity as a secure filesystem/backup medium
>   (fwd)

> Assume the existence of a for-profit service provider.
> The service provider needs to cover the costs of the service,

Which in general are a flat rate. I'll try to fill in the various issues you
raise with as real-world numbers as I can get.

> plus enough profit to make the effort interesting.
> The prices charged need to reflect the costs (or be higher)
> or the service provider can't make money and goes out of business.
> So what are the costs, and who can be talked into paying for them?

Everyone who uses the system. The question is who do we want to define as
'using'. I believe the provider of information is not 'using' the system
in the strictest sense but rather making an investment of their time and
effort in somebodies elses business model and hoping that work will bring
them profit.

> 1) Storage of information - Storage currently costs < $1/MB for raw disk, 
> and getting cheaper by the minute,

Assuming your 5 year plan, a 1G drive costs $100 or less. That's roughly
$20/G-yr.

> but sysadmins, lawyers, etc.
> cost money and they're not getting cheaper as fast.

Granted, but except for the sys-admin you don't need them often. In all the
years I have been in business I've needed a lawyer only a couple of times.
Say $1k/lawyer-year.

> The costs of the equipment for permanent storage are probably about
> 10-50% more than the costs for storing for 5 years;

A Cyrix P200 runs about $600 for a working box (I bought one just a few
weeks ago). So that gives us about $120/cpu-year.

> the costs of administration (assuming inflation is limited to
> some small number) can be covered by an annuity.

For reliable system administartion you're looking at 5 people (3 8-hr.
shifts plus weekend coverage of 2 12 hour shifts). A sys admin with the
skills and maturity to work in this environment is going to run you in
the neighborhood of $40-60k/yr. So this means we're looking at, on the
outside, $300k/sysadmin-yr. This is the real cost of doing business.

> Every technology upgrade or two you need to copy the archives to 
> new storage media, and data that doesn't get accessed often may
> get migrated out to slower or perhaps even offline media;
> storage contracts need to reflect that retrieving data that hasn't
> been accessed in a while may involve some delay.

Not shure what the normal ratio for online versus offline storage actualy
is currently, say 10:1.

> 2) Transmission of information - Roughly proportional to MB/time -

This seems overly simplistic to me. The actual cost of the bandwidth is
reasonably fixed for a given site. Also if we throw in the other servers
this model becomes a bit more flexible. Remember, under the Eternity model
we don't know *which* server is being hit for the request. This means we
have to consider the actual end user costs in a much more complicated
fashion. It's one of the issue I'm working on myself. It does not seem clear
cut nor well researched at all.

> unlike storage, this one's not predictable, unless the provider and
> author agree in advance (e.g. N free accesses per year, per password.)
> So the provider could charge the reader for access,

I fail to see the profit in giving away plans for man portable nukes or to
turn commen cooking yeast into a THC producing horde when the various
groups around the world would pay so many millions (or would that be
billions) for some to get it and some to keep others from getting it. The
potential for a access auction hasn't been explored as far as I am aware.

> or use advertising
> banners to fund retrieval costs (if that remains a valid model
> for financing the web over the next N years, especially if the
> readers retrieve data through anonymizers.)

I believe advertising would be a necessity. The question is how. Would the
payoff for doing a print ad in a magazine be worth it? Should it only
be advertised on the net?

> 3) Legal defense - This one's harder to predict :-)  

If the Eternity model works the question becomes which one and where are
they located? Whose legal defense? Which jurisdiction? Put up an offer for 1
weeks notice of any pending legal actions against the network and make it
worth the while. Say $500M. It would of course be fully anonymous and the
information would have to be testable prior to payout. Leaves two option.
First, the data is verified and the source is paid. Second, the server
goes down and the source isn't paid - thus ruining the reputation of a
server. It isn't clear to me what the total impact would be on the network
as a whole. One way to mediate this impact on a given server would be to
share the cost of the payout between the various servers. It is after all in
all the servers best interest to keep as many going unharrassed and as well
prepared and for-warned as possible.

> The current US climate is that service providers are relatively immune 
> as long as they cooperate with subpoenas and court orders,

True, but the current model doesn't allow anonymous providers as the norm
either. The users may be anonymous but the ISP isn't.

> If they want their names known, they can include them in the contents
> of the data that readers retrieve, independent of what the server does.

Then there is no reason to use an anonymous network, simply put the data on
their own webpage and sell it direct, cut out the middle-man a tried and
true business tactic.

> and accumulate reputation capital under those nyms.

One of the assumptions is that the source, individual server, and sink
are all anonymous to each other as well as Mallet. Now we're changing the
rules of the game in the middle of the game. Apples and oranges.

If I were actualy going to impliment something like this I wouldn't even
consider it unless I had at least $500k worth of funding for each of the
first 5 years. That start up cost, $2.5M, is what is going to kill it.



    ____________________________________________________________________
   |                                                                    |
   |       The most powerful passion in life is not love or hate,       |
   |       but the desire to edit somebody elses words.                 |
   |                                                                    |
   |                                  Sign in Ed Barsis' office         |
   |                                                                    | 
   |            _____                             The Armadillo Group   |
   |         ,::////;::-.                           Austin, Tx. USA     |
   |        /:'///// ``::>/|/                     http://www.ssz.com/   |
   |      .',  ||||    `/( e\                                           |
   |  -====~~mm-'`-```-mm --'-                         Jim Choate       |
   |                                                 ravage@ssz.com     |
   |                                                  512-451-7087      |
   |____________________________________________________________________|

At 09:26 AM 1/20/98 -0600, Jim Choate wrote:
>> From: Bill Stewart <bill.stewart@pobox.com>
>> 1) Storage of information - Storage currently costs < $1/MB for raw disk, 
>> and getting cheaper by the minute,
>
>Assuming your 5 year plan, a 1G drive costs $100 or less. That's roughly
>$20/G-yr.
>> The costs of the equipment for permanent storage are probably about
>> 10-50% more than the costs for storing for 5 years;
>A Cyrix P200 runs about $600 for a working box (I bought one just a few
>weeks ago). So that gives us about $120/cpu-year.

You missed my point - the cost of a CPU-year and GB-year drop rapidly
every year; the costs of storage plus associated CPU for 100 years of 
storage are probably just a bit higher than for 5 years.

>> but sysadmins, lawyers, etc.
>> cost money and they're not getting cheaper as fast.
>Granted, but except for the sys-admin you don't need them often. In all the
>years I have been in business I've needed a lawyer only a couple of times.
>Say $1k/lawyer-year.

For normal businesses, that's true, but eternity servers, if they can
be physically located, probably will require a bit more lawyery.

>> the costs of administration (assuming inflation is limited to
>> some small number) can be covered by an annuity.
>
>For reliable system administartion you're looking at 5 people (3 8-hr.
>shifts plus weekend coverage of 2 12 hour shifts). A sys admin with the
>skills and maturity to work in this environment is going to run you in
>the neighborhood of $40-60k/yr. So this means we're looking at, on the
>outside, $300k/sysadmin-yr. This is the real cost of doing business

You're probably right, though you may not need 7x24 sysadmins,
and they can share their time maintaining quite a lot of servers if
they're all cookie-cutter.

>Not shure what the normal ratio for online versus offline storage actualy
>is currently, say 10:1.

Somewhere between 10:1 and 100:1; the cost of sysadmins making tapes
may be higher than the media itself :-)

>> 2) Transmission of information - Roughly proportional to MB/time -
>This seems overly simplistic to me. The actual cost of the bandwidth is
>reasonably fixed for a given site. 

The price of pipes is very roughly proportional to their size.
The size of the pipes you need are roughly proportional to how much
data gets transmitted during your busy hour.  This gets cheaper
every year as well, but nowhere near as fast, and the volume
depends on how interesting the data is to potential readers
as well as on how big it is.  

> Remember, under the Eternity model
>we don't know *which* server is being hit for the request. 

There are different models for how to run an Eternity server,
but any server will need bandwidth based on how much is getting retrieved.
In any case, the machine transmitting data knows which request
it's responding to, even if it doesn't really know who the
requester or the author are, so it's not unreasonable to charge
a retrieval fee or stick on an advertising banner.

>> unlike storage, this one's not predictable, unless the provider and
>> author agree in advance (e.g. N free accesses per year, per password.)
>> So the provider could charge the reader for access,
>
>I fail to see the profit in giving away plans for man portable nukes or to
>turn commen cooking yeast into a THC producing horde when the various
>groups around the world would pay so many millions (or would that be
>billions) for some to get it and some to keep others from getting it. The
>potential for a access auction hasn't been explored as far as I am aware.

You might want to give away that THC recipe for free just to end
the drug war, or to make supplies cheap and plentiful, or because
it's a bogus recipe using your own special brand of yeast.  
It's the principle of the thing :-)  On the other hand,
if the provider of the service deliberately doesn't know what
the documents on the system are, to him it's just shipping bits,
and content is Somebody Else's Problem.

>> or use advertising
>> banners to fund retrieval costs (if that remains a valid model
>> for financing the web over the next N years, especially if the
>> readers retrieve data through anonymizers.)
>
>I believe advertising would be a necessity. The question is how. Would the
>payoff for doing a print ad in a magazine be worth it? Should it only
>be advertised on the net?

I'm not talking about how to get users - I'm talking about those
annoying banners from Doubleclick and LinkExchange that fund many
of the interesting services on the net because advertisers think
it's worth their money to buy impressions.

>>    If they want their names known, they can include them in the contents
>>    of the data that readers retrieve, independent of what the server does.
>Then there is no reason to use an anonymous network, simply put the data on
>their own webpage and sell it direct, cut out the middle-man a tried and
>true business tactic.
>>    and accumulate reputation capital under those nyms.
>One of the assumptions is that the source, individual server, and sink
>are all anonymous to each other as well as Mallet. Now we're changing the
>rules of the game in the middle of the game. Apples and oranges.

The assumption is that True Names and physical locations are probably
anonymous.  Doesn't mean that a reader is going to retrieve documents
of unknown content by unknown authors - you may very well know that
Zed WareMeister has the best deal on slightly-used Microsoft Products
and can be found on alt.eternity.warez, and his reputation capital 
accumulates under the pseudonym rather than his True Name.
But there will also be authors whose names are well-known but whose 
locations aren't - Salmon Rushdie may be selling digitally signed copies 
of his latest book online, but he still doesn't want to be easy to find.
				Thanks! 
					Bill
Bill Stewart, bill.stewart@pobox.com
PGP Fingerprint D454 E202 CBC8 40BF  3C85 B884 0ABE 4639