Use real byte size limits

[coot]

Fixes #1727.

[nfrisby]

This looks very reasonable to me. Let's discuss it at the Planning Meeting (briefly, I hope) on the 12th.

I wrote one comment that possibly suggests a non-trivial simplification.

Also, this PR's current logic boils down to

some mini protocol states will reject a message of more than 64 kilobytes and others will reject a message of more than 2,500,000 bytes.

Is that right? That seems fine for Cardano mainnet. But is it a fit for other (potential) users of this library? If not, then we ultimately need to make those numbers depend on eg the block type blk.

[coot]

Is that right? That seems fine for Cardano mainnet. But is it a fit for other (potential) users of this library? If not, then we ultimately need to make those numbers depend on eg the block type blk.

Block size is a parameter of the network and yes these limits should ultimately be configurable rather than hard coded. But block size is independent of block type (though tx size, probably does depend on tx type). So ultimately, these limits should be provided by a cardano-node configuration file (or hard coded there).

I think this makes it clear why we should keep the interface proposed in this PR. Note also timeouts are different, since they are much less conservative than size limits, but in long run we will also need to make them configurable.

[nfrisby]

Marcin and I reached consensus on a Consensus Planning Meeting call. Approving.

[karknu]

LGTM

[coot]

bors merge

[iohk-bors]

Build failed:

• ci/hydra:Cardano:ouroboros-network:required

[coot]

bors merge

[iohk-bors]

Build succeeded:

• buildkite/ouroboros-network
• ci/hydra:Cardano:ouroboros-network:required