How difficult is it to increase Bitcoin divisibility?

[rebuilder]

Currently, Bitcoins are divisible to 8 decimal places. It's conceivable that if Bitcoin gains real momentum, in the future that may not be enough granularity. It's not going to be a problem for a while, quite possibly ever,  but it may be in the future. So, some thinking on this follows:

I'll use the term "Satoshi", as previously coined by others, to mean the current smallest available Bitcoin unit.

-From what I understand, increasing the divisibility would require modifications to the protocol such that transfers of any sums below one Satoshi would not be compatible with clients lacking the modifications. Correct?

-I assume that transfers of sums up to 8 decimal places would still be possible between "new" clients and "old" ones, or at least the modifications could be done in such a way that this was the case.

Now, if it becomes necessary to add divisibility, how troublesome will this be? Let's say I'm selling nuts and bolts and I realize one Satoshi is too much for a bolt. I have a customer who really, really just wants to buy one bolt from me, but is only willing to pay .1 Satoshi for it, which I think is a fair price. So, I start using a modified client that allows for spending of sub-Satoshi amounts, and I tell my customer to do the same. Let's say I had 1000 Satoshi before, now I have 1000.1. I do many such transactions, until the smaller sums eventually form several Satoshi.

So, I suppose the client would be able to collate sub-Satoshi sums into Satoshis so that I can still send 1 Satoshi or more to someone not using a client capable of handling sub-Satoshi amounts. The other option is that sub-Satoshi transfers essentially "corrupt" my wallet in such a way that the BTC it contains are no longer compatible with legacy versions of the protocol. This is not the case, right?

Basically I'm wondering if there's any serious incompatibility that would result from an increase in divisibility, or if it's just a matter of needing to upgrade to a new client if one wishes to use sub-Satoshi amounts, without any added troubles if one does not. As far as I can tell, it should be pretty trouble-free, but is there something I'm missing? If increased divisibility amounted to forking the block chain, then maybe it should be done sooner rather than later.

[theymos]

The change is pretty simple, I think: The transaction and block versions would increase, and the sign bit of nValue would be treated as a sort of "enhanced precision" flag.

This is totally incompatible with old clients, though. The two networks won't be able to interact, since old clients will view all enhanced-precision transactions as being invalid. I don't think there's any way to do it in a backward-compatible way. Like all changes of this sort, it'll be phased in. This has been done several times before. The new rule applies only after some block number is reached, so everyone moves to the new rule simultaneously. Here is a past backward-incompatible change that was phased in, for example:

Code:

if (nBestHeight > 84000)
{
    if (nKeysCount > 20)
        return false;
    nOpCount += nKeysCount;
    if (nOpCount > 201)
        return false;
}

[stakhanov]

Shouldn't we make the change now, when it is still relatively simple?

[genjix]

Shouldn't we make the change now, when it is still relatively simple?

I agree. Either variable length or INT128.

The beginning 13 bits are not even used and are always 0 even. Maybe they could be used as a flag to indicate another INT64 following the same format on the end.