On 2/5/23 08:11, Richard Purdie wrote:
On Sat, 2023-02-04 at 15:47 -0600, Alex Stewart wrote:
On 2/3/23 17:06, Richard Purdie wrote:
On Fri, 2023-02-03 at 12:26 -0600, Alex Stewart wrote:
Hey Josh,
I have been roadmapping SBOM generation for NI's yocto distro and have a
few open questions about the future of SPDX and the create-spdx bbclass.
Since your name seems to be attached to both of those, I figure you
might have the best insight here.
Also posting to the OE-core ML so that this discussion can help other
members.
1. SPDX 3 timeline. I hear that SPDX 3 is going to be a complete rewrite
of the spec, with more support for modern SBOM discussion topics like
VEX and more comprehensive vulnerability tracking. And it also seems to
me that the timeline for its release is very behind schedule [1], but
still in active development. Can you give a SWAG for how close that new
spec is to completion? Are we months away or years?
2. If/when SPDX 3 support is released, is it to be assumed that the SPDX
facilities in OE core are going to be upgraded to handle it?
Assuming someone does the work, which is likely, yes. We did namespace
the class in master to prepare for that.
3. The rest of my org is interested in CycloneDX as our common SBOM
format. Have there been any discussions about supporting CDX SBOMs in
OE-core? Any blockers there; or is it something that my org could author
and upstream if we decide to go that route?
At this point OpenEmbedded/Yocto Project has decided to go the SPDX
route for various reasons.
Are those reasons documented somewhere?
Something about CDX rubs me the wrong way (besides it being named like
an off-brand printer company), but I can't put my finger on what. So if
there are technical reasons that it is less desirable for the OE
usecase, I'd like to know about them.
I think I share that same feeling but it is hard to put a finger on
why.
I'm not sure anything was documented but there was a discussion by the
TSC when we made the decision. Some rough random thoughts:
* SPDX did go the extra step of becoming an ISO standard
* We (as a community) have much better contacts with the SPDX
community. It is a fellow Linux Foundation project.
* We already were using SPDX license identifiers so this is a natural
progression/alignment.
* Looking at the CDX repositories and mailing lists, the discussions
and contributions do look to be from a much smaller ecosystem
* SPDX are interested in engaging with us, which as Joshua mentions,
does have benefits. If we run into challenges, we can likely seek
help. We're actively involved with 3.0 which means we can hopefully
ensure it works well for us.
Both specifications do encode roughly the same information so the
project alignment with SPDX and the "social" aspects likely tipped the
balance.
That's good context. Thanks.
My understanding is that CDX has better support for embedding
vulnerability (+VEX) and attestation elements into its DOM, which is
something that our Aero-Def customers will be interested in. I suppose I
can build workflows to add that information after converting the OE-SPDX
document to CDX, but I'd like to integrate the whole thing into an OE
build, if possible.
I think CDX added VEX information in the last year or so and SPDX plans
to do so in their 3.0. I have had the feeling it is due soon.
I'm not sure I want to see two formats being
added directly to the core, the better solution would likely be to
translate the SPDX output into CycloneDX if/as needed.
I'm concerned about the lossiness of that conversion. Based on the
CDX-SPDX mapping document in the cdx2spdx tool repo [1], they seem
roughly compatible. But I haven't been able to find a clean tool which
converts the other direction, nor a mapping document for the SPDX->CDX
pathway.
You could take that different ways but it could mean people aren't
finding a need to convert SPDX to CDX, meaning SPDX is working for
them.
From what I've seen, I think the SBOM ecosystem is too immature to say
whether one is preferred over the other. It seems like most folks are
just working with whatever falls out of their static dep analysis
tooling, or gets handed to them by their security auditor.
I haven't yet seen evidence that any of our customers have an SBOM
*pipeline* such that they would need to pick a common spec; but I
suspect that is coming.
I don't really want to have two partially implemented SBoM mechanisms
in YP/OE so I'd probably suggest any CDX code was a separate layer at
this point. If there is actual missing capability, I'd be asking SPDX
to resolve it :)
Understood. I'll assume that we'll have to maintain our own layer, if we
go that route.
Cheers,
Richard
(copying Kate Stewart to the discussion)
--
Alex Stewart
Software Engineer - NI Real-Time OS
NI (National Instruments)
alex.stew...@ni.com
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#176793):
https://lists.openembedded.org/g/openembedded-core/message/176793
Mute This Topic: https://lists.openembedded.org/mt/96729387/21656
Group Owner: openembedded-core+ow...@lists.openembedded.org
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub
[arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-
