This discussion sounds to be veering slightly off track.  I think we should be focusing on how we will ease the transition for new users to get on the network and use it.  Talking about the necessity and costs of running full nodes in the future is important, but irrelevant here:  unless we don't want users who aren't willing to run full nodes, we need to accommodate users who want to simply "use" the network, not necessarily "support" it.  I'm making an assumption here that we want new users whether they use a full node or not.  Greg's point looks like it's veering towards "we don't want to grow the network unless we're going to get more full nodes out of it."     I'm of the opinion, like Mike Hearn, that the number of full nodes needed for a healthy network is not O(N) in the number of users of the network.  I expect it to be something more like O(sqrt(N))... or perhaps there's even an upper limit above which the network gets no benefit, even if all 7 billion humans were using it.  (the bottleneck would be size of blocks and CPU processing power at that point, not a shortage of full nodes).  Would we rather have a system that is "full-node-or-nothing" and drive away users that won't support the network, or accommodate those users with various gradations of participation?

I believe my proposal for an address-based meta-chain (or something like it) is exactly what is needed in the long run.  It could almost obsolete this entire discussion.  However, as Greg pointed out there is a long, treacherous path between the theory I presented, and a working&robust implementation that can serve as the backbone for future SPV nodes.  I have every intention to help pioneer that when Armory has other major features completed (such as multi-sig), but it's not something that we can even consider in the near- or medium-term as a solution to rely on.  I'd be surprised if any such solution existed in the next 6-12 months. 

I think it is very much in everyone's interest here to encourage new users to start "using" Bitcoin, even if they don't "support" it.  As long as there is a convenient channel for interested users to get more information about the system, the benefits of spending the effort to run a full node, and the features available in more-advanced clients that they might benefit from, then I'm not personally concerned about a shortage of full nodes, and we should carry forward with the idea of promoting SPV nodes for the really-new users.

-Alan




On Tue, Dec 4, 2012 at 4:41 PM, Gregory Maxwell <gmaxwell@gmail.com> wrote:
On Tue, Dec 4, 2012 at 3:58 PM, Mike Hearn <mike@plan99.net> wrote:
>> It sounds to me that you're insisting that you're asking people who
>> oppose degrading our recommendations to commit to a costly rushed
>> development timeline. I think this is a false choice.
>
> Hardly. I don't have any particular timeline in mind. But I disagree
> we have "forever". New ideas have a certain time window to take off
> and become credible.

Marketing initiatives have limited windows.  This matters, perhaps,
when you're some VC pumping cash into a startup with the hopes of
being the next stockmarket pump and dump darling.  Outside of that
people use whatever they use because it works for them.

And by the numbers Linux desktops are more common than they've ever
been— and certainly Linux kernel _systems_ half the people I know have
one in their pocket and its hard to go more than a few hours without
touching one.  To some extent the "Year of the Linux desktop" is a bit
like the "Year of being able to turn lead into gold" ... we can turn
lead into gold now, but the particle accelerators, atomic power, and
atomic weapons enabled by the same technology are far more interesting
due to the particle realities of this. So we didn't get the ubiquitous
Linux desktop: We got the ubiquitious Linux server, the ubiquitous
Linux-kernel smart phone, the ubiquitous Linux television, media
player, HVAC controller, etc. instead.

Desktops— well, that didn't meet people's hopes though I think not for
the lack of marketing on the part of Linux, but because Apple stepped
up and produced middle ground products that attracted a larger
audience. Especially as MSFT dropped the ball. They did some things
better, had a running start, and had a non open source software
business model which made reaping rewards easier.

But I don't see how any of this has anything to do with Bitcoin...
Except for the point that if Bitcoin doesn't become the money system
everyone uses and instead becomes the money system infrastructure all
the systems people use depend on— just as Linux has with the desktop,
where it might not be on the desktop but its in router firmware, cloud
servers, and just about everything else— I wouldn't consider that much
of a loss.

> time window, eventually people just give up and move on. Does anyone
> take desktop Linux seriously anymore? No. "The year of desktop Linux"
> is a joke. People took it seriously in 2001 but despite great progress
> since, the excitement and attention has gone. There were steady
> improvements over the last 10 years but nobody is creating desktop
> Linux startups anymore

Bitcoin already missed its first— and perhaps only— fad window in any
case. Today people say "Bitcoin? Thats still around? I thought it got
hacked". ... thanks to compromised centralized services.

> It's unclear we need to have every man and his dog run a full node.

Every man and his dog? Perhaps not.  But as many as can— probably so.

If we depend on the organic need for full nodes to overcome cost and
effort to run one there will always be major incentives to let someone
else do that, and the system would have its equilibrium right on the
brink of insecurity. Perhaps worse, since insecurity is most obvious
retrospectively. Security doesn't make for a good market force.

> Tor is a successful P2P network where the number of users vastly
> outstrips the number of nodes, and exit nodes in particular are a
> scarce resource run by people who know what they're doing and commit
> to it.

Tor is a distributed but controlled, by a small number of directory
authority operators, system.

It is a good system. But it has a trust model which is categorically
weaker than the one in Bitcoin.  If you want something where a
majority of a dozen signing keys— hopefully in the hands of trusted
parties— can decide the state of the system you can produce someting
far superior to Bitcoin— something that gives near instant
non-reversable transactions, something that gives good client security
without the complexity of a SPV node, etc.

But that isn't Bitcoin.

> Even with no incentives, they were able to obtain
> the resources they need.

And yet every tor user— if the have the bandwidth available can be a
full internal relay and the software nags them to do it (and also nags
them to act as invisible bridges for blocking avoidance), and every
user is technically able to run an exit (though they don't bludgeon
users to do that, because of the legal/political/technical issues
involved).  To do any of this doesn't require a user to switch to
different software, and the tor project has previously opposed client
only software.

> So why should Bitcoin be different?

It's less different than you make it out to be— but it _is_ different.
  Bitcoin is a distributed currency. The value of bitcoin comes from
the soundness of its properties and from the persistence of its
security. If the integrity of the distributed ledger is disrupted the
damage produced, both in funds stolen and in undermining the
confidence of the system, can be irreversible. Because Bitcoin's value
comes from confidence in Bitcoin and not from the specific
functionality of Bitcoins (they're random numbers that sit on your
disk) even if the ledger isn't actually compromised but people
reasonably believe it could be compromised that undermines the value.
 Tor, on the other hand, is a functioning system whos value depends on
its current usefulness, and not the past or future security.

Compare in your mind— Say everyone just found out that at block
420,000 Bitcoin would stop enforcing signature correctness or block
subsidy values (and this wasn't going to be fixed), and you also found
out that one year from now Tor would hand over their sites, source
code repositories, and directory authority keys to Iran (and you have
no suspicion that they already had done so).   How fast would you stop
using Tor vs how fast would to sell whatever coins you could?

> We can easily send a clear and consistent "this is important, please
> help" message without complicated auto-upgrade/downgrade schemes that
> risk annoying users.

I don't think we really can send such a message.  Thanks just the same
as asking for donations, not completely unsuccessful but not easy to
make successful either.  You're arguing for people running distinct
software which has no capability to be a full node, and changing what
they're doing in order to support the network. This maximizes the
cost, because in addition to the real cost the user must take a
switching cost too, and deemphasizes investing in keeping the full
node software as usable because 'oh just run a lite node if the full
is too slow'.

------------------------------------------------------------------------------
LogMeIn Rescue: Anywhere, Anytime Remote support for IT. Free Trial
Remotely access PCs and mobile devices and provide instant support
Improve your efficiency, and focus on delivering more value-add services
Discover what IT Professionals Know. Rescue delivers
http://p.sf.net/sfu/logmein_12329d2d
_______________________________________________
Bitcoin-development mailing list
Bitcoin-development@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bitcoin-development