So I've created 2 new repositories with changed rules regarding
sequencenumbers:

https://github.com/maaku/bitcoin/tree/sequencenumbers2

This repository inverts (un-inverts?) the sequence number. nSequence=1
means 1 block relative lock-height. nSequence=LOCKTIME_THRESHOLD means 1
second relative lock-height. nSequence>=0x80000000 (most significant bit
set) is not interpreted as a relative lock-time.

https://github.com/maaku/bitcoin/tree/sequencenumbers3

This repository not only inverts the sequence number, but also interprets
it as a fixed-point number. This allows up to 5 year relative lock times
using blocks as units, and saves 12 low-order bits for future use. Or, up
to about 2 year relative lock times using seconds as units, and saves 4
bits for future use without second-level granularity. More bits could be
recovered from time-based locktimes by choosing a higher granularity (a
soft-fork change if done correctly).

On Tue, Aug 25, 2015 at 3:08 PM, Mark Friedenbach <mark@friedenbach.org>
wrote:

> To follow up on this, let's say that you want to be able to have up to 1
> year relative lock-times. This choice is somewhat arbitrary and what I
> would like some input on, but I'll come back to this point.
>
>  * 1 bit is necessary to enable/disable relative lock-time.
>
>  * 1 bit is necessary to indicate whether seconds vs blocks as the unit of
> measurement.
>
>  * 1 year of time with 1-second granularity requires 25 bits. However
> since blocks occur at approximately 10 minute intervals on average, having
> a relative lock-time significantly less than this interval doesn't make
> much sense. A granularity of 256 seconds would be greater than the Nyquist
> frequency and requires only 17 bits.
>
>  * 1 year of blocks with 1-block granularity requires 16 bits.
>
> So time-based relative lock time requires about 19 bits, and block-based
> relative lock-time requires about 18 bits. That leaves 13 or 14 bits for
> other uses.
>
> Assuming a maximum of 1-year relative lock-times. But what is an
> appropriate maximum to choose? The use cases I have considered have only
> had lock times on the order of a few days to a month or so. However I would
> feel uncomfortable going less than a year for a hard maximum, and am having
> trouble thinking of any use case that would require more than a year of
> lock-time. Can anyone else think of a use case that requires >1yr relative
> lock-time?
>
> TL;DR
>
> On Sun, Aug 23, 2015 at 7:37 PM, Mark Friedenbach <mark@friedenbach.org>
> wrote:
>
>> A power of 2 would be far more efficient here. The key question is how
>> long of a relative block time do you need? Figure out what the maximum
>> should be ( I don't know what that would be, any ideas?) and then see how
>> many bits you have left over.
>> On Aug 23, 2015 7:23 PM, "Jorge Timón" <
>> bitcoin-dev@lists.linuxfoundation.org> wrote:
>>
>>> On Mon, Aug 24, 2015 at 3:01 AM, Gregory Maxwell via bitcoin-dev
>>> <bitcoin-dev@lists.linuxfoundation.org> wrote:
>>> > Seperately, to Mark and Btcdrank: Adding an extra wrinkel to the
>>> > discussion has any thought been given to represent one block with more
>>> > than one increment?  This would leave additional space for future
>>> > signaling, or allow, for example, higher resolution numbers for a
>>> > sharechain commitement.
>>>
>>> No, I don't think anybody thought about this. I just explained this to
>>> Pieter using "for example, 10 instead of 1".
>>> He suggested 600 increments so that it is more similar to timestamps.
>>> _______________________________________________
>>> bitcoin-dev mailing list
>>> bitcoin-dev@lists.linuxfoundation.org
>>> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
>>>
>>
>