fc438a2f44eda02a

refactor: switch hashing from blake2s to sha256 by default

f2fcbf7f485987b6

fix(storage-proofs): spec sync

The first round of updating naming and adjusting some smaller things better matching the spec.

Important to note
- fixed logic in tapering, such that it starts at the last layer note the first
- split labeling proofs and encoding proofs
- add missing round of labeling proofs for the last layer
- add missing checks for column proof
- improve efficiency of the comm_r check

7cca60a02cbf9c95

feat: add script to pin params and publish dnslink (#909)

Adds pin-params.sh which will pin the parameters to cluster and publish the cid to proofs.filecoin.io

To keep things simple the script assumes you'll run it against a dir with just the current version of params in.

It keeps the previous versions of the params by adding a link called `prev` to the new params dir. So you get

.
├── v13-proof-of...meta
├── v13-proof-of...params
...
├── prev
    ├── v12-proof-of...meta
    ├── v12-proof-of...params
    ...

and so on.

The scripts aims to be idempotent... if you re-run it on the same params it wont endlessly link the same version of the parms.

You need to provide the following env vars

- CLUSTER_TOKEN - the basic auth user:pass string to pin things to cluster.ipfs.io
- DNSIMPLE_TOKEN - the api key to update the dnslink for proofs.filecoin.io

You need the following tools available on your PATH

- ipfs-cluster-ctl - to talk to cluster. download from https://dist.ipfs.io/#ipfs-cluster-ctl
- npx - to run dnslink-dnsimple. comes free with npm these days
- ipfs - you probably have one of those.

The script will take a while to run. That's expected...
It may redundently re-add the prams to your local ipfs, but thats ~1m on my machine.
Pinning the new params on cluster can take a while. Depends on the prevailing network winds.
Re-running it on the same params is pretty quick, though somewhat redundant.

It's fast and loose with its use of `cut`, so it'll need a little more smarts when your version numbers hit 100.

The storage space needed to store all the params will become an issue for the gateways pretty quickly...
we have about 100GB free per node which is supposed to be used to temporarily cache general gateway content. I'll work on a solution for that.

Cluster has about 40TB of space, so fill yer boots in that regard.

License: MIT
Signed-off-by: Oli Evans <[email protected]>

330cdf9053027e93

feat: publish params@v16

c6a4dfa8d0f2a1a4

refactor(filecoin-proofs): add more information to piece size errors

* fix(comment): remove dead comment
* fix(errors): better error message
* feat(errors): add more error messaging
* refactor(errors): alert user to EOF and 0-byte read

6d318b061441cd73

feat(storage-proofs): revive drgporep (#937)

feat(storage-proofs):  revive drgporep

a7170e5b5b56ea52

expose piece-writing function which expects caller to bin pack (#935)

* feat(pieces): add write_and_preprocess to API

- comment the behavior of the add_piece function
- add write_and_preprocess function, which will ultimately replace add_piece

* fix(typo): fix typo in comment

* feat(commp): generate comm_p while writing and preprocessing

66297323d138fc2f

feat(bellman): use new bellman version which does not panic (#936)

c6e7436d2c113ea6

fix: remove unneeded restriction on the seal seed (#934)

fix: remove unneeded restriction on the seal seed

4409cb32f0e0e9ef

Merge pull request #933 from filecoin-project/feat/compute-comm-d

Add compute_comm_d to API.