Quote from: tvbcof on June 21, 2015, 06:12:48 PM

Quote from: molecular on June 21, 2015, 03:24:49 PM

Quote from: Odalv on June 21, 2015, 02:28:08 PM

Yes, you are right. But solex is presenting IBTL like it reduces bandwidth to 1/32

That's probably a misunderstanding. He said: "Obviously the real-time unconfirmed transaction overhead does remain. and is large long-term.".

Solex seems to have been envisioning IBLT's to be akin to how RAID obtains 'magic' results by use of parity bits. He would be wrong about this...but...

...actually a Merkle tree structure offers the ability to do some pretty significant optimizations and 'magic'. If, for instance, a branch can be cut off because it contains some taint, all of the twigs and leaves beyond it are irrelevant. Pruning could be arbitrarily effective at discriminating and reducing bandwidth depending on the taint ratio. 1/32 is hardly inconceivable and it could be much more.

The key element if IBLT would be a structured ordering of transactions within a block. My personal feeling is that this is like the difference in handling ability between water in liquid form and water in solid form. An ice cube is much more flexible to transport, and much easier to match to a mold (e.g., an identifiable ice tray configuration), but it is also more brittle. An ice cube is also much easier to catch in a strainer than it's liquid counterpart.

that is the intent.