bmdsoftware / dicoogle-ipfs-storage Goto Github PK

Summary

Heads up:

• IPFS ecosystem will switch to CIDv1 as the default in the near future (https://github.com/ipfs/ipfs/issues/337)
• The way CID is extracted from ipfs:// addresses in this project will break when it attempts to read a CIDv1
• The fix is to use CID library instead of Multihash when parsing ipfs:// addresses

Problem

This code assumes every CID is CIDv0 (looks the same as Multihash, so it is parsed as Multihash here):

                    public InputStream getInputStream() throws IOException {
                        Multihash filePointer = Multihash.fromBase58(location.toString().replaceFirst("ipfs://", ""));

                        byte[] fileContents = new byte[0];
                        try {
                            fileContents = ipfs.cat(filePointer);
                        } catch (IOException e) {
                            logger.error("Failed to retrieve object", e);
                            return null;
                        }

Try parsing ipfs://bafybeigdyrzt5sfp7udm7hu76uh7y26nf3efuylqabf3oclgtqy55fbzdi – it is not Base58 (it is CIDv1 in Base32), so will throw an exception.

Solution

Always assume localtion is a CID and parse it using a library that understands CIDv1.

You can use CID implementation from java-cid, you can also peek at how CIDs are used in Peergos

See:

• difference between CIDv0 (Multihash in Base58bct) and CIDv1 (multibase prefix+version+codec+multihash)

Happy to answer any questions / guide

cc @rlebre for awareness