CIP-19: Shwap

[Wondertan]

• Rendered
• Discussion

[jcstein]

Missing required section to add:

• Rationale

Optional sections:

• Test Cases

[Wondertan]

I decided to merge the whole spec into CIP instead of maintaining a separate doc. Having it externally seems cumbersome and prone to breaks. We would have to cross-link various required sections from CIP to the actual step. Instead, I will link celestia-node specs to CIP. In the worst case, we will copy the documents if that does not work for some reason.

[Wondertan]

After spending a few hours, I was able to nicely integrate the CIP into specs, so there is no need to do any copying

[jcstein]

sharing this on github as well, I think it would be good if we align this CIP and the discussion from rsmt2d . based on what i can tell, it already is

[Wondertan]

Connecting previous discussion from celestiaorg/celestia-node#3205

[jcstein]

left some comments and committable suggestions

[ebuchman]

Just a first pass to clarify some things

[ebuchman]

is this 16 bit limit enforced elsewhere in the core protocol?

[Wondertan]

Not sure I understand the question.

[ebuchman]

Oh I mean, is there a limit on how many rows there can be enforced somewhere else in the protocol?

[Wondertan]

Ah not that I would be aware of. The u16 is chosen because it maxes at 2TB squares with 512shares, which is more than enough. Might be worth clarifying that as well in the spec.

[Wondertan]

cc @rootulp, do we have any row number limits defined in core?

[rootulp]

Yea! The max row number is equivalent to 2 * the max effective square size because the row number seems applicable to the extended data square. The max effective square size is currently 64 but it could be bumped to 128 via a governance proposal.

Related table comparing original data square size and extended data square size: https://gist.github.com/rootulp/bbf10f6e9cf114816aaa994eb64b63a4

[jcstein]

leaving this discussion open to migrate into the forum

[rootulp]

Re-reviewed the first half.

[jcstein]

approved, but before merging:

• 2 remaining comments should first be resolved
• latest changes need to be pulled in, and CIP-19 added to readme

[jcstein]

needs to be linted

[jcstein]

the only remaining step is to link these 3 lines to the correct serialization heading, as there are now 2 different headings. before, there were 2 headings with the same content: "serialization"

cips/cip-19.md:154:1 MD051/link-fragments Link fragments should be valid [Context: "[Serialized](#serialization)"]
cips/cip-19.md:198:1 MD051/link-fragments Link fragments should be valid [Context: "[Serialized](#serialization)"]
cips/cip-19.md:256:1 MD051/link-fragments Link fragments should be valid [Context: "[Serialized](#serialization)"]

additional context, we cannot have a link that goes back to 2 headings. it just will not work. how would a link direct you to a heading, if there are 2 identical headings?