The OpenChain Project held a mini-summit adjacent to the Linux Foundation Open Source Summit North America. Check out our opening keynote for some substantial data points on our project, our standards for license compliance and security assurance, and the type of support you can get with adoption.
We continued with a presentation from our board member Helio (CARIAD), with a strong focus on how people can use automation in the practical implementation of important compliance and security processes at scale.
The final presentation drilled further down the stack, and we had a great contribution from the LG Electronics team as their explained FOSSLight, an open source tool for open source compliance or security management with sophisticated dashboard and automation. This solution is gaining traction in South Korea and is well worth attention globally.
The overarching event this year had around 2,000 physical attendees and 2,000 virtual, and we were delighted to welcome some new faces to our corner of the open source community. It was also a pleasure to see many familiar faces in the room.
Minutes Prepared By Steve Kilbane of Analog Devices
- Expecting the Security Spec to graduate from ISO/IEC at end of July.
- Shane has produced 8 case studies using ChatGPT.
- Helio on “State of Tooling in Open Source Automation” (Helio can probably share his slides, if they’re not already on the LF platform)
- Tools, Trends, Insights.
- Previous trend was license compliance.
- Current trend is security.
- Few can consume SBOMs.
- Lots of gaps for license compliance automation.
- We need open data, avoiding control of that data by one entity.
- Binary analysis will displace source-only scans.
- I think this point here is that, current binary scans aren’t sufficient, but as we move up SLSA levels, we’ll have more attestations from the build, and those will be sufficient.
- Poor data quality, especially vulnerability databases.
- PURLs prevent vendor lock-in to a given DB.
- We need unique identifiers for software.
- We need to share the data of package review and curation, but need to overcome concerns from legal departments.
- Should we share scanner output first? (ahead of curations?)
- We should try to fix upstream (to have better compliance info / metadata)
- Helio wants data to be standardised; I was unclear whether Helio was saying data should be centralised or de-centralised (sorry, Helio). I wasn’t clear whether the call was for a federated network of standard servers.
- Licensing isn’t the same as security. Lots in common, but different use-cases, with different audiences, so have different docs to explain your systems and tools.
- License compatibility: Multiple tools / matrices in use, but they’re all legally subjective and dependent on jurisdiction.
- Snippet matching
- V. expensive in terms of time (and, therefore, money)
- Weirdly, Helio argued that Synopsys has given up on Snippet matching, as they’ve all but abandoned Protex. Hub has snippet-matching – we use it all the time at ADI.
- Suggests that ChatGPT et al. will make snippet matching more relevant and useless, at the same time, because it’ll generate new boilerplate from everyone’s code.
- Note to self: Look into MatchCode, which Helio mentioned.
- SBOMs
- Not good, don’t have all the data.
- Often can’t read them anyway.
- Tools do not integrate them well.
- SBOMs need to be validated – but even a valid SBOM can contain junk data, if the data is wrong in the first place.
- Collaboration opportunities
- “Live inventory of FOSS tools and their capabilities” – which sounds like the capability map / tooling landscape the OpenChain Automation WG was working on last year.
- FossLight presentation from LG (fosslight.org)
- Scans with ScanOSS and ScanCode.
- Bunch of package managers supported.
- Has a built-in workflow – SBOM management?
- Has a Jenkins CI for the prechecker.
- Mails vulnerability notices to the dev team.
- Has a Supply Chain Management section, for third-party code.
- Unclear how many of the features being mentioned are part of the OSS product, and how many are still internal-only for LG.
- I didn’t spot where the clearing/curation decision feeds back into a later scan.
- Sounds like developers can only upload single packages at a time to be scanned; bulk upload is an internal-only package at the moment.
- Shane mentioned a cautionary tale on automation from a Chinese company. They asked their OSPO to set up Fossology and (some other tool I didn’t catch). The OSPO budgeted three hours to do the job. They spent a week on it, then gave up and bought Black Duck. So we have a way to go on making tooling easier to set up.
