Removed ring dependency #127

Xynnn007 · 2022-09-13T07:21:39Z

Summary

Due to #27. This pr removed ring crate, and use RustCrypto crates instead.

Release Note

replaced ring crate with RustCrypto
refactored CosignVerificationKey implementation and interface
fixed the code which references CosignVerificationKey
added RSA keys and signing algorithms support (RSA PSS & RSA PKCS#1 v1.5)

API Changes

Changed all the from_XXX functions for CosignVerificationKey's SignatureDigestAlgorithm parameter into SigningScheme
Deleted the enum SignatureDigestAlgorithm
Replace struct CosignVerificationKey into enum.

Documentation

flavio

Overall I like the change.

There are however two main things that have to be addressed:

Be able to use RSA keys to verify signatures
Given a PEM encoded key, have the code automatically figure out the encryption and signing algorithm. Or provide the user a helper function that does that

flavio · 2022-09-13T09:47:22Z

src/crypto/signing_key/ed25519.rs

@@ -399,16 +398,16 @@ mod tests {
        let pubkey = key
            .public_key_to_pem()
            .expect("export private key to PEM format failed.");
+        println!("{}", pubkey);


nit, probably a leftover?

Yea, sry for that. It was used for test.

flavio · 2022-09-13T12:21:35Z

src/crypto/verification_key.rs

-            verification_algorithm,
-            data,
-        })
+        Self::from_der(key_pem.contents.as_slice(), signing_scheme)


I miss the usability of the old implementation. Now, as a user I would have to specify the signing scheme when importing the key. The signing scheme is made by encryption algorithm + digest algorithm.

The old code leveraged a functionality of ring that provided back the type of encryption algorithm being used; see lines 69-95.

Thanks to that, the verify example was able to handle different kind of public keys without having the user to specify the encryption type.

Moreover, the current code lost the ability to handle RSA keys. I happen to have some container images that have been signed with a RSA key. This change breaks their verification.

In case you are interested:

curl -O https://ftp.suse.com/pub/projects/security/keys/container–key.pem cargo run --example verify -- -k ./container–key.pem registry.suse.com/suse/sle-micro/5.0/toolbox

This will fail with the following error:

2022-09-13T12:21:07.095762Z INFO sigstore::cosign::client_builder: Rekor public key not provided. Rekor integration disabled 2022-09-13T12:21:07.095804Z INFO sigstore::cosign::client_builder: No Fulcio cert has been provided. Fulcio integration disabled Image verification failed: Cannot create public key verifier: Pkcs8 spki error : Ecdsa-P256 from der bytes to public key failed: unknown/unsupported algorithm OID: 1.2.840.10045.2.1

FWIW, the next release of the rsa crate will include some pretty major overhauls of signature support, including dedicated RSASSA-PSS and RSASSA-PKCS#1v1.5 signing/verifying keys which use the traits from the signature crate and hopefully support for things like parsing PSS params via PKCS#1/PKCS#8.

The current low-level signing APIs will also continue to work with some minor changes as well.

Thanks for your example, it will be very helpful.

I miss the usability of the old implementation. Now, as a user I would have to specify the signing scheme when importing the key. The signing scheme is made by encryption algorithm + digest algorithm.

The old code leveraged a functionality of ring that provided back the type of encryption algorithm being used; see lines 69-95.

In fact, the padding scheme used in the old implementation of RSA-related schemes only included PSS, so the API could only receive one parameter DigestAlgorithm. There are other padding schemes, like OAEP, PKCS#1v1.5. Thus EC keys may not share the same from_pem/der API with RSA keys.

Also, the public key file usually only contains information of the key itself. For example an EC public key only contains information of its type (ec public key, by OID) and curve (by OID), but not its signing usage.

So I think it is hard to do that, (TBH, I think the old implementation for the API is not reasonable)

Moreover, the current code lost the ability to handle RSA keys. I happen to have some container images that have been signed with a RSA key. This change breaks their verification.

Maybe we can delay this pr. When the RSA key-interface is done, I'll rebase and go on doing this.

FWIW, the next release of the rsa crate will include some pretty major overhauls of signature support, including dedicated RSASSA-PSS and RSASSA-PKCS#1v1.5 signing/verifying keys which use the traits from the signature crate and hopefully support for things like parsing PSS params via PKCS#1/PKCS#8.

The current low-level signing APIs will also continue to work with some minor changes as well.

Thanks for the information! I wonder when the next release will be published. :)

We're trying to wrap it up soon. This is probably the last major PR we'd like to get merged first prior to another release: RustCrypto/RSA#183

Ok, I got your idea about the developer convenience. I added two interfaces try_from_der() and try_from_pem(). The two interfaces receive only one parameter data: &[u8], and will decide the verification algorithm (digest, padding ,...) with a default value. For example, by default ECDSA_P256_SHA256_ASN1 for a ec-p256 public key and RSA_PSS_2048_8192_SHA256 for a RSA public key (in future)
Does this make sense?

Go ahead with the SignatureDigestAlgorithm refactoring, they are pretty nice

What do you mean? Should SignatureDigestAlgorithm be kept or deleted? Sry I can not understand this ;)

Ok, I got your idea about the developer convenience. I added two interfaces try_from_der() and try_from_pem(). The two interfaces receive only one parameter data: &[u8], and will decide the verification algorithm (digest, padding ,...) with a default value. For example, by default ECDSA_P256_SHA256_ASN1 for a ec-p256 public key and RSA_PSS_2048_8192_SHA256 for a RSA public key (in future)
Does this make sense?

Yes, it does make sense. I like the UX

What do you mean? Should SignatureDigestAlgorithm be kept or deleted? Sry I can not understand this ;)

My fault, I like the cleaning/refactoring done by this PR. That's the way to go 👍

I like the PR. If possible, I would like RSA support to be in before merging it. Otherwise, we can merge it in the main branch, but not tag a new release until this is added back.

Thanks. Let me finish RSA support in this PR then.

New rsa crate prerelease here:

https://docs.rs/rsa/0.7.0-rc.0/

It includes new PSS-specific SigningKey/VerifyingKey types:

https://docs.rs/rsa/0.7.0-rc.0/rsa/pss/struct.SigningKey.html

It'd be good to give us feedback on anything that might require a breaking change soon, as we intend to cut a final release in the next week or so.

FYI, for support for PKCS#1/PKCS#8 PSS keys, this is a tracking issue specifically for supporting them:

RustCrypto/RSA#177

flavio

I like the PR. If possible, I would like RSA support to be in before merging it. Otherwise, we can merge it in the main branch, but not tag a new release until this is added back.

flavio · 2022-09-16T06:58:54Z

src/crypto/verification_key.rs

-            verification_algorithm,
-            data,
-        })
+        Self::from_der(key_pem.contents.as_slice(), signing_scheme)


Ok, I got your idea about the developer convenience. I added two interfaces try_from_der() and try_from_pem(). The two interfaces receive only one parameter data: &[u8], and will decide the verification algorithm (digest, padding ,...) with a default value. For example, by default ECDSA_P256_SHA256_ASN1 for a ec-p256 public key and RSA_PSS_2048_8192_SHA256 for a RSA public key (in future)
Does this make sense?

Yes, it does make sense. I like the UX

What do you mean? Should SignatureDigestAlgorithm be kept or deleted? Sry I can not understand this ;)

My fault, I like the cleaning/refactoring done by this PR. That's the way to go 👍

Xynnn007 · 2022-09-18T13:32:20Z

Please do not review the newly push-ed commit now, because it uses a modified version of rsa crate. I've made a PR for the change. After the PR is merged I'll change this commit.

lukehinds

great work thanks @Xynnn007

Xynnn007 · 2022-09-19T08:58:27Z

great work thanks @Xynnn007

Hi @lukehinds Thanks for your approval.

However there is some significant things to clarify. This pr now involves non-official crypto crate version and must not be merged until the upstream of rsa merges the pr (The current version does not provide Clone and Debug trait implementation for signing keys).

I pushed the second commit to collect suggestions about the rsa key length. There are two design schemes:

Add a member usize for RSA-related signing schemes enums to specify the key length (which is implemented in the commit). It is convenient to generate new keypair, but may seem weird when call RSAKeyPair::to_verification_key() to carry a usize.
Another choice is to set a default value of the key size when generating the RSAKeyPair (like 2048 or 4096 bits). This may sacrifice the flexibility, but can avoid confusing users and avoid letting users to set it manually.

Please share some views about which one is better, thanks.

cc @flavio @tarcieri

lukehinds · 2022-09-20T08:28:50Z

Got you @Xynnn007 you have the dependency pinned to your fork until upstream changes merge. Thanks for pointing that out, I will change this to a hold.

Regarding the two options. I am more inclined to allow users to pass a usize in. How about we allow users to set key length, yet if the key length is None, we default to 4096?

lukehinds

hold until RSA upstream patch is merged.

tarcieri · 2022-09-20T13:50:15Z

Note: the RSA patch is merged

Xynnn007 · 2022-09-20T15:19:55Z

Note: the RSA patch is merged

Great, thanks! When the new version of rsa is published I will import it, instead of {git = "xxx", rev = "xxx"}

lukehinds · 2022-09-29T10:55:54Z

@tarcieri do you have any insight into when v0.7.0 moves from rc to a release?

flavio · 2022-09-29T11:49:33Z

We can already consume the rc from crates.io: https://crates.io/crates/rsa/0.7.0-rc.0

lukehinds · 2022-09-29T12:20:39Z

We can already consume the rc from crates.io: https://crates.io/crates/rsa/0.7.0-rc.0

Is this sufficient @Xynnn007 ?

flavio · 2022-09-29T12:27:02Z

I tried building using the released crate and the code doesn't compile.

Aside from that, I tried the current code (the one that builds)::

curl -O https://ftp.suse.com/pub/projects/security/keys/container–key.pem
cargo run --example verify -- -k ./container–key.pem registry.suse.com/suse/sle-micro/5.0/toolbox

and the verification fails. I didn't look too much into that, maybe @Xynnn007 has a better idea about why this RSA key is not working anymore 🤔

Xynnn007 · 2022-09-29T13:52:33Z

Yea, I pushed a pr on 0.7.0-rc.0 to help, so we should wait for the newest tag after 0.7.0-rc.0
@lukehinds @flavio

tarcieri · 2022-09-29T13:52:43Z

@lukehinds we'd like to get a final release out soon, although it'd be good to know if there are any remaining issues.

lukehinds · 2022-09-30T11:22:29Z

I think we should merge this, and hold on the release for 0.7.0 , at which point we can just update cargo.toml, that way it will save folks managing rebases when this later lands. We can cut the release then.

How does that sound @Xynnn007 ?

Xynnn007 · 2022-09-30T11:42:29Z

I think we should merge this, and hold on the release for 0.7.0 , at which point we can just update cargo.toml, that way it will save folks managing rebases when this later lands. We can cut the release then.

How does that sound @Xynnn007 ?

Ok, I've just change the crate ref. Also, current implementation may not be perfect as I see, and we can make it better over time.

flavio · 2022-09-30T11:54:50Z

@Xynnn007 Please, before merging it, can you figure out what is wrong with the RSA signature I've pointed out here?

This is a regression

Xynnn007 · 2022-09-30T12:00:29Z

@Xynnn007 Please, before merging it, can you figure out what is wrong with the RSA signature I've pointed out here?

This is a regression

Ah, please wait for some time. Let me do this

flavio · 2022-09-30T12:07:47Z

Ah, please wait for some time. Let me do this

Sure, no hurries! 😄

I just wanted to ensure this PR wouldn't get accidentally merged with this open issue.

Xynnn007 · 2022-09-30T12:12:00Z

@flavio Hey, there is a trick. The image you've given is sign by RSA_PKCS1_SHA256, and I set the default signing scheme of RSA is RSA_PSS_SHA256, because the padding scheme PSS is safer than PKCS1

You can use an extra parameter to tell the signing scheme

cargo run --example verify -- -k ./container–key.pem --signing-scheme RSA_PKCS1_SHA256 registry.suse.com/suse/sle-micro/5.0/toolbox

And will result

    Finished dev [unoptimized + debuginfo] target(s) in 0.14s
     Running `target/debug/examples/verify -k './container–key.pem' --signing-scheme RSA_PKCS1_SHA256 registry.suse.com/suse/sle-micro/5.0/toolbox`
2022-09-30T12:08:50.238132Z  INFO sigstore::cosign::client_builder: Rekor public key not provided. Rekor integration disabled
2022-09-30T12:08:50.238193Z  INFO sigstore::cosign::client_builder: No Fulcio cert has been provided. Fulcio integration disabled
Image successfully verified

Still hold on, please. I found some bugs for notes, let me fix it in a few minutes.

Xynnn007 · 2022-09-30T12:17:10Z

Done, please take a look whether there is problems left
@flavio @lukehinds

lukehinds · 2022-09-30T12:48:38Z

could you rebase with #136 pls

Xynnn007 · 2022-09-30T13:45:57Z

@lukehinds hey done

lukehinds · 2022-09-30T14:27:30Z

thanks @Xynnn007

@flavio is your comment still applicable?

lukehinds · 2022-10-05T07:19:00Z

I should have held off 0.5.0 for this one, my bad. @flavio as above, are you OK with this one to merge?

Cargo.toml

flavio · 2022-10-05T16:03:13Z

@flavio Hey, there is a trick. The image you've given is sign by RSA_PKCS1_SHA256, and I set the default signing scheme of RSA is RSA_PSS_SHA256, because the padding scheme PSS is safer than PKCS1

You can use an extra parameter to tell the signing scheme

cargo run --example verify -- -k ./container–key.pem --signing-scheme RSA_PKCS1_SHA256 registry.suse.com/suse/sle-micro/5.0/toolbox

And will result

    Finished dev [unoptimized + debuginfo] target(s) in 0.14s
     Running `target/debug/examples/verify -k './container–key.pem' --signing-scheme RSA_PKCS1_SHA256 registry.suse.com/suse/sle-micro/5.0/toolbox`
2022-09-30T12:08:50.238132Z  INFO sigstore::cosign::client_builder: Rekor public key not provided. Rekor integration disabled
2022-09-30T12:08:50.238193Z  INFO sigstore::cosign::client_builder: No Fulcio cert has been provided. Fulcio integration disabled
Image successfully verified

Please, forgive me if I bring this topic up again. Is there a way to understand if a RSA key is using the pkcs v1 1.5 or the pss signature algorithm? I tried searching for it, but I couldn't find anything about that. Maybe @Xynnn007 or @tarcieri know about that...

I'm so obsessed by this topic because I would like the consumers of the sigstore crate to be able to build tools that behave like cosign. With cosign, I don't have to specify what kind of algorithms are used by an input key, it will try to guess it using a list of supported algorithms.

Now, this PR is doing something in this direction too with the PublicKeyVerifier::try_from method. It would be great to extend this method to cover also RSA_PKCS1 verification keys.

tarcieri · 2022-10-05T22:34:49Z

Is there a way to understand if a RSA key is using the pkcs v1 1.5 or the pss signature algorithm?

I'm not 100% clear on this, but I believe PKCS#1 v2.2 as defined in RFC8017 makes it possible to mark a key as intended for use with PSS, and also provide the additional PSS parameters (e.g. hash function, MGF, and salt length) needed to instantiate the algorithm. See:

https://datatracker.ietf.org/doc/html/rfc8017#appendix-A.2.3

Note that this isn't yet supported by the pkcs1 crate. We have a tracking issue for it here: RustCrypto/formats#694

tarcieri · 2022-10-05T23:19:33Z

I released rsa v0.7.0-rc.1 which may fix the compile errors.

We also have a tracking issue for a final v0.7.0 release here: RustCrypto/RSA#205

flavio · 2022-10-06T07:59:43Z

I released rsa v0.7.0-rc.1 which may fix the compile errors.

Yes, that fixes the compilation errors. Thanks!

I've updated this PR to make use of this pre-release from crates.io instead of obtaining the code from the git repository

- replace ring crate with RustCrypto - refactor CosignVerificationKey implementation and interface - fix the code which references CosignVerificationKey Signed-off-by: Xynnn007 <mading.ma@alibaba-inc.com>

Xynnn007 · 2022-10-06T08:33:34Z

I released rsa v0.7.0-rc.1 which may fix the compile errors.

We also have a tracking issue for a final v0.7.0 release here: RustCrypto/RSA#205

Great, thanks

I released rsa v0.7.0-rc.1 which may fix the compile errors.

Yes, that fixes the compilation errors. Thanks!

I've updated this PR to make use of this pre-release from crates.io instead of obtaining the code from the git repository

Hi, I'll rebase the newest version and fix the errors.

Now, this PR is doing something in this direction too with the PublicKeyVerifier::try_from method. It would be great to extend this method to cover also RSA_PKCS1 verification keys.

Hi agreed. However as @tarcieri said, the related functions of RSA Crypto is still not prepared. I think we can wait for the functions to be prepared and then make a new pr to fix this. Before that, we still should need to manually set the algorithms for the key. Does it make sense? @flavio

src/crypto/verification_key.rs

flavio · 2022-10-06T13:36:54Z

Hi agreed. However as @tarcieri said, the related functions of RSA Crypto is still not prepared. I think we can wait for the functions to be prepared and then make a new pr to fix this. Before that, we still should need to manually set the algorithms for the key. Does it make sense? @flavio

I've looked at how cosign works. Currently users who want to use keys not generated by cosign must import them, see this document.

The cosign import-key-pair does NOT support the pps padding, I did a quick test with the latest code built from the main branch. Also, looking at the verify code, the RSA key is expected to be using pkcs11.

Because of that, I propose to change the code that guesses the RSA key to give priority to what cosign supports.
That translates to do the following change:

diff --git a/src/crypto/verification_key.rs b/src/crypto/verification_key.rs
index 93845e5897..2a0c470724 100644
--- a/src/crypto/verification_key.rs
+++ b/src/crypto/verification_key.rs
@@ -161,7 +161,9 @@ impl CosignVerificationKey {
                 ed25519bytes.as_ref(),
             )?))
         } else if let Ok(rsapk) = rsa::RsaPublicKey::from_public_key_der(der_data) {
-            Ok(Self::RSA_PSS_SHA256(pss::VerifyingKey::new(rsapk)))
+            Ok(Self::RSA_PKCS1_SHA256(
+                pkcs1v15::VerifyingKey::new_with_prefix(rsapk),
+            ))
         } else {
             Err(SigstoreError::InvalidKeyFormat {
                 error: "Failed to parse the public key.".to_string(),

I think that gives the UX I'm looking for :)

Close sigstore#27 Signed-off-by: Xynnn007 <mading.ma@alibaba-inc.com>

flavio

Thanks a lot @Xynnn007 for the awesome work and for your patience during the review process! 👏 ❤️

@bobcallaway

Enhancements ============ * update user-agent value to be specific to sigstore-rs (sigstore#122) * remove /api/v1/version from client by (sigstore#121) * crate async fulcio client (sigstore#132) * Removed ring dependency (sigstore#127) Others ====== * Update dependencies * Refactoring and examples for key interface (sigstore#123) * Fix doc test failures (sigstore#136) Contributors ============ * Bob Callaway (@bobcallaway) * Bob McWhirter (@bobmcwhirter) * Flavio Castelli (@flavio) * Luke Hinds (@lukehinds) * Xynnn (@Xynnn007) Signed-off-by: Flavio Castelli <fcastelli@suse.com>

tarcieri · 2022-10-10T18:17:48Z

FYI rsa v0.7.0 has been released: https://crates.io/crates/rsa/0.7.0

Xynnn007 · 2022-10-11T02:43:48Z

FYI rsa v0.7.0 has been released: https://crates.io/crates/rsa/0.7.0

Yea, thanks!

Xynnn007 force-pushed the refactor-remove-ring branch 2 times, most recently from fe13148 to 553f90e Compare September 13, 2022 08:46

lukehinds requested review from flavio and lukehinds September 13, 2022 11:09

flavio reviewed Sep 13, 2022

View reviewed changes

Xynnn007 force-pushed the refactor-remove-ring branch 3 times, most recently from b160821 to 76b1b9d Compare September 14, 2022 02:33

flavio reviewed Sep 16, 2022

View reviewed changes

Xynnn007 force-pushed the refactor-remove-ring branch from a5974ac to 460de25 Compare September 18, 2022 14:03

lukehinds approved these changes Sep 19, 2022

View reviewed changes

lukehinds requested changes Sep 20, 2022

View reviewed changes

flavio previously approved these changes Sep 29, 2022

View reviewed changes

Xynnn007 dismissed flavio’s stale review via 534cf4e September 30, 2022 11:41

Xynnn007 force-pushed the refactor-remove-ring branch from 460de25 to 534cf4e Compare September 30, 2022 11:41

lukehinds previously approved these changes Sep 30, 2022

View reviewed changes

Xynnn007 dismissed lukehinds’s stale review via 447c89f September 30, 2022 12:16

Xynnn007 force-pushed the refactor-remove-ring branch from 534cf4e to 447c89f Compare September 30, 2022 12:16

Xynnn007 force-pushed the refactor-remove-ring branch from 447c89f to 5b17f87 Compare September 30, 2022 13:00

flavio reviewed Oct 5, 2022

View reviewed changes

Cargo.toml Outdated Show resolved Hide resolved

Xynnn007 force-pushed the refactor-remove-ring branch from 6f266b7 to 4afde59 Compare October 6, 2022 08:17

refactor: remove ring dependency

237c6d8

- replace ring crate with RustCrypto - refactor CosignVerificationKey implementation and interface - fix the code which references CosignVerificationKey Signed-off-by: Xynnn007 <mading.ma@alibaba-inc.com>

flavio reviewed Oct 6, 2022

View reviewed changes

src/crypto/verification_key.rs Show resolved Hide resolved

Xynnn007 force-pushed the refactor-remove-ring branch from 4afde59 to 21b4e5c Compare October 6, 2022 14:43

feat: add rsa key module

5c14f24

Close sigstore#27 Signed-off-by: Xynnn007 <mading.ma@alibaba-inc.com>

Xynnn007 force-pushed the refactor-remove-ring branch from 21b4e5c to 5c14f24 Compare October 6, 2022 15:04

flavio approved these changes Oct 6, 2022

View reviewed changes

flavio merged commit a0ebf6b into sigstore:main Oct 7, 2022

Xynnn007 deleted the refactor-remove-ring branch October 8, 2022 01:22

Removed ring dependency #127

Removed ring dependency #127

Conversation

Xynnn007 commented Sep 13, 2022 • edited

Summary

Release Note

API Changes

Documentation

flavio left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Xynnn007 Sep 13, 2022 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Xynnn007 Sep 14, 2022 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tarcieri Sep 16, 2022 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

flavio left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Xynnn007 commented Sep 18, 2022

lukehinds left a comment

Choose a reason for hiding this comment

Xynnn007 commented Sep 19, 2022 • edited

lukehinds commented Sep 20, 2022

lukehinds left a comment

Choose a reason for hiding this comment

tarcieri commented Sep 20, 2022

Xynnn007 commented Sep 20, 2022

lukehinds commented Sep 29, 2022

flavio commented Sep 29, 2022

lukehinds commented Sep 29, 2022

flavio commented Sep 29, 2022

Xynnn007 commented Sep 29, 2022

tarcieri commented Sep 29, 2022

lukehinds commented Sep 30, 2022 • edited

Xynnn007 commented Sep 30, 2022

flavio commented Sep 30, 2022 • edited

Xynnn007 commented Sep 30, 2022

flavio commented Sep 30, 2022

Xynnn007 commented Sep 30, 2022

Xynnn007 commented Sep 30, 2022

lukehinds commented Sep 30, 2022

Xynnn007 commented Sep 30, 2022

lukehinds commented Sep 30, 2022

lukehinds commented Oct 5, 2022

flavio commented Oct 5, 2022

tarcieri commented Oct 5, 2022 • edited

tarcieri commented Oct 5, 2022

flavio commented Oct 6, 2022

Xynnn007 commented Oct 6, 2022

flavio commented Oct 6, 2022

flavio left a comment

Choose a reason for hiding this comment

tarcieri commented Oct 10, 2022

Xynnn007 commented Oct 11, 2022

Xynnn007 commented Sep 13, 2022 •

edited

Xynnn007 Sep 13, 2022 •

edited

Xynnn007 Sep 14, 2022 •

edited

tarcieri Sep 16, 2022 •

edited

Xynnn007 commented Sep 19, 2022 •

edited

lukehinds commented Sep 30, 2022 •

edited

flavio commented Sep 30, 2022 •

edited

tarcieri commented Oct 5, 2022 •

edited