Re: [bitcoin-dev] BIP 158 Flexibility and Filter Size

[Olaoluwa Osuntokun <laolu32@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 5952 bytes --]

> An example of that cost is you arguing against specifying and supporting
the
> design that is closer to one that would be softforked, which increases the
> time until we can make these filters secure because it
> slows convergence on the design of what would get committed

Agreed, since the commitment is just flat out better, and also also less
code to validate compared to the cross p2p validation, the filter should be
as close to the committed version. This way, wallet and other apps don't
need to modify their logic in X months when the commitment is rolled out.

> Great point, but it should probably exclude coinbase OP_RETURN output.
> This would exclude the current BIP141 style commitment and likely any
> other.

Definitely. I chatted offline with sipa recently, and he suggested this as
well. Upside is that the filters will get even smaller, and also the first
filter type becomes even more of a "barebones" wallet filter. If folks
reaally want to also search OP_RETURN in the filter (as no widely deployed
applications I know of really use it), then an additional filter type can be
added in the future. It would need to be special cased to filter out the
commitment itself.

Alright, color me convinced! I'll further edit my open BIP 158 PR to:

  * exclude all OP_RETURN
  * switch to prev scripts instead of outpoints
  * update the test vectors to include the prev scripts from blocks in
    addition to the block itself

-- Laolu


On Sat, Jun 9, 2018 at 8:45 AM Gregory Maxwell <greg@xiph.org> wrote:

> > So what's the cost in using
> > the current filter (as it lets the client verify the filter if they want
> to,
>
> An example of that cost is you arguing against specifying and
> supporting the design that is closer to one that would be softforked,
> which increases the time until we can make these filters secure
> because it slows convergence on the design of what would get
> committed.
>
> >> I don't agree at all, and I can't see why you say so.
> >
> > Sure it doesn't _have_ to, but from my PoV as "adding more commitments"
> is
> > on the top of every developers wish list for additions to Bitcoin, it
> would
> > make sense to coordinate on an "ultimate" extensible commitment once,
> rather
> > than special case a bunch of distinct commitments. I can see arguments
> for
> > either really.
>
> We have an extensible commitment style via BIP141 already. I don't see
> why this in particular demands a new one.
>
> >   1. The current filter format (even moving to prevouts) cannot be
> committed
> >      in this fashion as it indexes each of the coinbase output scripts.
> This
> >      creates a circular dependency: the commitment is modified by the
> >      filter,
>
> Great point, but it should probably exclude coinbase OP_RETURN output.
> This would exclude the current BIP141 style commitment and likely any
> other.
>
> Should I start a new thread on excluding all OP_RETURN outputs from
> BIP-158 filters for all transactions? -- they can't be spent, so
> including them just pollutes the filters.
>
> >   2. Since the coinbase transaction is the first in a block, it has the
> >      longest merkle proof path. As a result, it may be several hundred
> bytes
> >      (and grows with future capacity increases) to present a proof to the
>
> If 384 bytes is a concern, isn't 3840 bytes (the filter size
> difference is in this ballpark) _much_ more of a concern?  Path to the
> coinbase transaction increases only logarithmically so further
> capacity increases are unlikely to matter much, but the filter size
> increases linearly and so it should be much more of a concern.
>
> > In regards to the second item above, what do you think of the old Tier
> Nolan
> > proposal [1] to create a "constant" sized proof for future commitments by
> > constraining the size of the block and placing the commitments within the
> > last few transactions in the block?
>
> I think it's a fairly ugly hack. esp since it requires that mining
> template code be able to stuff the block if they just don't know
> enough actual transactions-- which means having a pool of spendable
> outputs in order to mine, managing private keys, etc... it also
> requires downstream software not tinker with the transaction count
> (which I wish it didn't but as of today it does). A factor of two
> difference in capacity-- if you constrain to get the smallest possible
> proof-- is pretty stark, optimal txn selection with this cardinality
> constraint would be pretty weird. etc.
>
> If the community considers tree depth for proofs like that to be such
> a concern to take on technical debt for that structure, we should
> probably be thinking about more drastic (incompatible) changes... but
> I don't think it's actually that interesting.
>
> > I don't think its fair to compare those that wish to implement this
> proposal
> > (and actually do the validation) to the legacy SPV software that to my
> > knowledge is all but abandoned. The project I work on that seeks to
> deploy
>
> Yes, maybe it isn't.  But then that just means we don't have good
> information.
>
> When a lot of people were choosing electrum over SPV wallets when
> those SPV wallets weren't abandoned, sync time was frequently cited as
> an actual reason. BIP158 makes that worse, not better.   So while I'm
> hopeful, I'm also somewhat sceptical.  Certainly things that reduce
> the size of the 158 filters make them seem more likely to be a success
> to me.
>
> > too difficult to implement "full" validation, as they're bitcoin
> developers
> > with quite a bit of experience.
>
> ::shrugs:: Above you're also arguing against fetching down to the
> coinbase transaction to save a couple hundred bytes a block, which
> makes it impossible to validate a half dozen other things (including
> as mentioned in the other threads depth fidelity of returned proofs).
> There are a lot of reasons why things don't get implemented other than
> experience! :)
>

[-- Attachment #2: Type: text/html, Size: 7041 bytes --]