What is the upper limit on block time?

[mczarnek]

Everyone is talking about shorter blocks.. I'd like to make longer blocks.  As long as I can make them while still being secure.

Would making a blockchain that has the difficulty target 1 week be too long for any reason?

I'm thinking that maybe issues with miners getting off and online would cause problems?  Is there someway to use smaller blocks but throw away all of them except for the ones that are one week apart?

[altcoinex]

There would be so many problems I don't know where to start...

[Foxpup]

Would making a blockchain that has the difficultly target 1 week be too long for any reason?

Too long for what? For an interplanetary currency with miners as far away as Eris? Sure, one week sounds like a good target. Maybe even two. Call it ten Earth days. The speed-of-light delay would make anything shorter than that infeasible. However, we don't have an interplanetary economy, and nobody here on Earth is willing to wait weeks for confirmations, so I'm not sure what you're trying to accomplish.

I'm thinking that maybe issues with miners getting off and online would cause problems?

No. Why would it?

Is there someway to use smaller blocks but throw away all of them except for the ones that are one week apart?

No. Timestamps can be faked. Only the block hash can determine if a block is valid. This means that you will occasionally get blocks very close together, resulting in orphaned blocks from miners on distant planets. This is unavoidable and happens with Bitcoin, too (except for the part about distant planets, obviously).

[mczarnek]

Would making a blockchain that has the difficultly target 1 week be too long for any reason?

Too long for what? For an interplanetary currency with miners as far away as Eris? Sure, one week sounds like a good target. Maybe even two. Call it ten Earth days. The speed-of-light delay would make anything shorter than that infeasible. However, we don't have an interplanetary economy, and nobody here on Earth is willing to wait weeks for confirmations, so I'm not sure what you're trying to accomplish.

I'm thinking that maybe issues with miners getting off and online would cause problems?

No. Why would it?

Is there someway to use smaller blocks but throw away all of them except for the ones that are one week apart?

No. Timestamps can be faked. Only the block hash can determine if a block is valid. This means that you will occasionally get blocks very close together, resulting in orphaned blocks from miners on distant planets. This is unavoidable and happens with Bitcoin, too (except for the part about distant planets, obviously).

Trying to make something other than a currency that has a decentralized checkpoint once a week.. I was wondering about fake blocks because it'd be nice if there was not huge variance of a few weeks or even months to find a block sometimes.

[TierNolan]

The rule for stability is that the block time has to be significantly longer than the latency within the network.  A solar system currency with a node on Pluto would have a latency of around 10 hours (Pluto to Neptune).

That means that you would need the block time to be at least 100-200 hours (4 - 8 days).

That could still work, as long as it was only used for planet to planet transfers.  Local currencies would be needed for each planet.

[spin]

Everyone is talking about shorter blocks.. I'd like to make longer blocks.  As long as I can make them while still being secure.

Would making a blockchain that has the difficulty target 1 week be too long for any reason?

I'm thinking that maybe issues with miners getting off and online would cause problems?  Is there someway to use smaller blocks but throw away all of them except for the ones that are one week apart?

With a long target variability also increases.  Using my rusty stats:
Assuming your hash rate is in steady state and your difficulty is stable you can use the exponential distribution with lambda=1/7 (7 days)
See: http://en.wikipedia.org/wiki/Exponential_distribution

We can work out using http://en.wikipedia.org/wiki/Exponential_distribution#Quantiles that:
99% of your blocks would be found within 32.2 days.  Which means 1 in 100 blocks will take longer than that.

If your use case is not constrained by latency you should probably check that your 99% block time is acceptable.  

In in bitcoins case using lambda =1 / 10 minutes we get 99% of blocks to be found in 46mins. Someone should check my stats though.  I'm rusty.

[DeboraMeeks]

Transactions would be stored in the mempool for that week, which means a node would have to store lots and lots of transactions temporarily. It would be messy!