You are viewing a plain text version of this content. The canonical link for it is here.
Posted to general@incubator.apache.org by tison <wa...@gmail.com> on 2026/09/07 23:30:05 UTC

[DISCUSS] Better License Verification Method

Hi PJ,

I recorded your vote.

To keep the vote thread focused and keep this specific idea discussion
focused, I am starting a new thread.

Let's see how we can improve the license/release verification method.

Comment inline.

> -1 binding
>
> I would prefer that only podlings that have done a clean release in
> the past can take advantage of expedited releases. I'm not against
> some mechanism to support shorter release turnarounds but I can't
> support the current proposal.

Technically speaking, the concurrent PPMC and IPMC vote is not an
expedited release as defined in [1]. It is still a formal and normal
vote by either the PPMC or the IPMC.

[1] https://www.apache.org/legal/release-policy.html#expedited-releases

> Or the release checklist should be expanded to include items like
> whether the source licenses have been validated - current wording
> typically just asks that the LICENSE and NOTICE are in the release.

I totally agree that we can extend the release criteria to help
podlings make better releases. This item, IMO, can also apply to TLPs.

Actually, the Incubator defines release guidelines in [2], and the
IPMC vote (release) SHOULD still follow the release policy defined in
[3], which states:

> Every ASF release MUST comply with ASF licensing policy. This requirement is of utmost importance and an audit SHOULD be performed before any full release is created. In particular, every artifact distributed MUST contain only appropriately licensed code per Apache Licensing Policy [4].

[2] https://incubator.apache.org/policy/incubation.html#releases
[3] https://www.apache.org/legal/release-policy.html#licensing
[4] https://www.apache.org/legal/resolved

So I think it is actually included. The wording could be improved, and
proposals are welcome.

But I can try to understand your concerns and consider whether we have
better methods to validate source licenses. So far, RAT can do
something but it would miss others that are not mechanically wrong.
ScanOSS can do some other things but it can still hardly replace human
checks.

What are human checks? They are defined as PPMC/IPMC votes cast on the
voting thread.

So technically, even if we repeat the requirement in the guidelines,
it doesn't change the practices.

The significant part is to enable and help PPMC/IPMC members verify
source licenses more effectively. And I find that the current AI agent
can reduce the workload (but you should not outsource the work).

My prompt is in Chinese, but it's as simple as asking the agent to
verify all the sources by following our policy, and it would search
the Internet to find similar sources. You can even ask for a lower
false-negative output in trade-off for a higher false-positive rate,
since false positives can be resolved by RM's answers without
canceling the vote. As examples, some recent -1 votes [5][6] are
spotted with the help of an agent.

[5] https://lists.apache.org/thread/bwr06m0mczd69tnq7h9opm1hnwhn4o9h
[6] https://lists.apache.org/thread/8tg6lwh154f5fxvjjkrpdlnxd9l9n37y

If you have other good practices or methods, feel free to share.

Our goal should be to help podlings release better, and waiting a few
more hours doesn't magically achieve that. Finally, we need a
combination of better methods and better education so that (P)PMC can
follow the policy more proactively and more easily.

> Convenience Binaries and their licenses/notices are not always considered.

I'm not sure what you want to emphasize here. Could you elaborate a bit?

Best,
tison.

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by tison <wa...@gmail.com>.
> it's as simple as asking the agent to
> verify all the sources by following our policy, and it would search
> the Internet to find similar sources.

Some outcomes look like [1][2] and result in [3][4].

[1] https://github.com/apache/asyncband/pull/275
[2] https://github.com/apache/asyncband/pull/279
[3] https://github.com/apache/asyncband/blob/27c30674/LICENSE#L203-L450
[4] https://github.com/apache/asyncband/blob/27c30674/RELEASE.md?plain=1#L136

This, however, cannot replace human verification and doesn't confirm
the absence of a license violation, just as a passed test doesn't mean
the software has no bugs. But new methods help raise the quality
baseline, where license compliance is part of an ASF release's
quality.

In my experience handling license compliance in companies, it's
typically grouped with security checks. You can create some analogies
and do comparisons to get a new perspective on handling issues spotted
by either side.

Best,
tison.

tison <wa...@gmail.com> 于2026年9月8日周二 07:37写道:
>
> > IPMC vote (release) SHOULD still follow the release policy defined in
>
> As this uses the uppercase form, it is a MUST.
>
> Best,
> tison.
>
> tison <wa...@gmail.com> 于2026年9月8日周二 07:30写道:
> >
> > Hi PJ,
> >
> > I recorded your vote.
> >
> > To keep the vote thread focused and keep this specific idea discussion
> > focused, I am starting a new thread.
> >
> > Let's see how we can improve the license/release verification method.
> >
> > Comment inline.
> >
> > > -1 binding
> > >
> > > I would prefer that only podlings that have done a clean release in
> > > the past can take advantage of expedited releases. I'm not against
> > > some mechanism to support shorter release turnarounds but I can't
> > > support the current proposal.
> >
> > Technically speaking, the concurrent PPMC and IPMC vote is not an
> > expedited release as defined in [1]. It is still a formal and normal
> > vote by either the PPMC or the IPMC.
> >
> > [1] https://www.apache.org/legal/release-policy.html#expedited-releases
> >
> > > Or the release checklist should be expanded to include items like
> > > whether the source licenses have been validated - current wording
> > > typically just asks that the LICENSE and NOTICE are in the release.
> >
> > I totally agree that we can extend the release criteria to help
> > podlings make better releases. This item, IMO, can also apply to TLPs.
> >
> > Actually, the Incubator defines release guidelines in [2], and the
> > IPMC vote (release) SHOULD still follow the release policy defined in
> > [3], which states:
> >
> > > Every ASF release MUST comply with ASF licensing policy. This requirement is of utmost importance and an audit SHOULD be performed before any full release is created. In particular, every artifact distributed MUST contain only appropriately licensed code per Apache Licensing Policy [4].
> >
> > [2] https://incubator.apache.org/policy/incubation.html#releases
> > [3] https://www.apache.org/legal/release-policy.html#licensing
> > [4] https://www.apache.org/legal/resolved
> >
> > So I think it is actually included. The wording could be improved, and
> > proposals are welcome.
> >
> > But I can try to understand your concerns and consider whether we have
> > better methods to validate source licenses. So far, RAT can do
> > something but it would miss others that are not mechanically wrong.
> > ScanOSS can do some other things but it can still hardly replace human
> > checks.
> >
> > What are human checks? They are defined as PPMC/IPMC votes cast on the
> > voting thread.
> >
> > So technically, even if we repeat the requirement in the guidelines,
> > it doesn't change the practices.
> >
> > The significant part is to enable and help PPMC/IPMC members verify
> > source licenses more effectively. And I find that the current AI agent
> > can reduce the workload (but you should not outsource the work).
> >
> > My prompt is in Chinese, but it's as simple as asking the agent to
> > verify all the sources by following our policy, and it would search
> > the Internet to find similar sources. You can even ask for a lower
> > false-negative output in trade-off for a higher false-positive rate,
> > since false positives can be resolved by RM's answers without
> > canceling the vote. As examples, some recent -1 votes [5][6] are
> > spotted with the help of an agent.
> >
> > [5] https://lists.apache.org/thread/bwr06m0mczd69tnq7h9opm1hnwhn4o9h
> > [6] https://lists.apache.org/thread/8tg6lwh154f5fxvjjkrpdlnxd9l9n37y
> >
> > If you have other good practices or methods, feel free to share.
> >
> > Our goal should be to help podlings release better, and waiting a few
> > more hours doesn't magically achieve that. Finally, we need a
> > combination of better methods and better education so that (P)PMC can
> > follow the policy more proactively and more easily.
> >
> > > Convenience Binaries and their licenses/notices are not always considered.
> >
> > I'm not sure what you want to emphasize here. Could you elaborate a bit?
> >
> > Best,
> > tison.

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Julian Hyde <jh...@gmail.com>.
I don’t think any name is needed. Let’s just call it a release.

We’ve simply removed the restriction that the IPMC vote couldn’t start until the PPMC vote ended.

It is my hope that podling release managers will use their discretion and, most of the time, wait until the PPMC vote has completed. This is an escape handle for cases like where, on the final day of the IPMC vote for the 4th release candidate, someone notices a typo in the NOTICE file.  

Since I’m one of the people who asked for a vote here, I should say: I see this change in the rules as a “small reversible step”. If it leads to better outcomes (faster graduation of better prepared TLPs, less toil on mentors and IPMC members) then let’s make more small, reversible changes. If it doesn’t, let’s back it out and try something else.

Julian


> On Sep 7, 2026, at 5:32 PM, Greg Stein <gs...@gmail.com> wrote:
> 
> On Mon, Sep 7, 2026 at 7:05 PM Dave Fisher <wa...@apache.org> wrote:
>> ...
> 
>> Incubator podlings simply have no policy route to an “Expedited Release”.
>> None, allowing that would be a new discussion. Expedited Releases must
>> occur on private@ and require informing the board afterwards.
>> 
>> If we need a name then how is “Compressed Release”?
>> 
> 
> Maybe call it an "Incubator-Approved Release"? We're talking about the IPMC
> (the formal PMC constructed per the Bylaws) approving a release. What more
> needs to be done?
> 
> Historically, anybody can make a release once they get (3) +1 votes on that
> release. I see no reason to treat the IPMC any differently from the other
> 200 PMCs at the Foundation, just because the repository being released has
> been labeled as a "podling". The Foundation does not recognize that
> construct in its Bylaws, nor any Resolutions I'm aware of, which formalize
> the podling concept as an independent entity.
> 
> I believe Tison is correct here, in saying that (3) IPMC Members may vote
> for a podling release to be made, before waiting for a PPMC vote.
> 
> Not sure on the name. As all releases are "Incubator-Approved".
> "Compressed" seems fine, but doesn't quite land properly. It also sounds
> like the "faster than 72h" rule or somesuch. "Overlapped Approval" or
> something? Blue-skying.
> 
> So needless to say, I asked Grok. It provided a great naming and policy
> guidance:
> 
> --
> 
> Keep the release name ordinary (Apache Foo 1.2.3). Name the *procedure*:
> 
> A release vote may close as *passed by concurrence* when the first-stage
> thread has both (a) a passing first-body tally and (b) at least three
> binding +1 votes from the second body. In that case the second-stage vote
> is not opened.
> 
> RESULT subject:
> 
> [RESULT][VOTE] Release Apache Foo 1.2.3-rc2 — passed by concurrence
> 
> Body line:
> 
> Binding second-body votes on this thread: John, Jane, Tom. Second-stage
> vote waived.
> 
> --
> 
>> ...
> 
> Cheers,
> -g


---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Hongtao Gao <ha...@apache.org>.
Hi Tison, Stefan, Paul,

SkyWalking Eyes [1] is one existing option. We use it in SkyWalking’s
CI to check license headers and dependency licenses.

For these well-defined checks, I prefer rule-based tools: their
behavior is more predictable and reproducible, with explicit rules and
clear pass/fail results. That makes them a reliable CI baseline.

Agents can help draft the rules and configuration, which are often the
trickiest parts to get right.

[1] https://github.com/apache/skywalking-eyes

Best,
Hongtao


On Wed, Sep 9, 2026 at 2:19 PM Paul King <pa...@asert.com.au> wrote:
>
> On Wed, Sep 9, 2026 at 3:30 PM tison <wa...@gmail.com> wrote:
> >
> > > To me this is all automateble. Why can't ASF have a blessed & centrally
> > > maintained way to check different file types at least as it pertains to
> > > licenses?
> >
> > It would be more than generous if you could give some concrete ideas
> > or a (referred) implementation.
>
> I find the various Gradle license plugins combined with a PolicyMCP
> goes a long way. Justin did various versions of a Python PolicyMCP
> implementation, I mostly use GroovyPolicyMCP (I have both normal
> Groovy and native versions, the latter faster than I can get combining
> Python/Rust).
>
> This is the kind of detail it gave on a recent Ivy release vote:
>
> **The jQuery Treeview plugin isn't listed in LICENSE.** Both zips
> bundle `jquery.treeview.js` (`asciidoc/js/` in src, `doc/js/` in bin).
> It's Treeview 1.2, copyright Jörn Zaefferer and Myles Angell,
> dual-licensed MIT/GPL — a separate work from jQuery core. The
> `LICENSE` file has a section for jQuery (John Resig) but nothing for
> Treeview, so it doesn't fully account for the package's exact content.
> The fix is a short "License for jQuery Treeview plugin" section taking
> the MIT option.
>
> One example from a recent Commons Collections release:
>
> 2. commons-collections4-4.6.0-javadoc.jar has no META-INF/LICENSE.txt or
>    NOTICE.txt, unlike the other jars. It carries the JDK's legal/ directory
>    instead, where legal/LICENSE is the GPLv2 text (with Classpath Exception
>    via ASSEMBLY_EXCEPTION). Relatedly, the binary distribution bundles
>    jQuery 3.7.1 and jQuery UI (MIT) under apidocs/script-dir/, which the
>    root LICENSE.txt does not mention, though apidocs/legal/jquery.md and
>    jqueryUI.md are present. This is inherited from stock javadoc and is
>    common across ASF Java projects; the source release is unaffected.
>
> These happen to be using Claude but you can use your favourite agent.
>
> Paul.
>
> > Best,
> > tison.
> >
> > Stefan Krawczyk <st...@gmail.com> 于2026年9月9日周三 13:21写道:
> > >
> > > To me this is all automateble. Why can't ASF have a blessed & centrally
> > > maintained way to check different file types at least as it pertains to
> > > licenses? I think the majority of this can be 100% deterministic code
> > > without the need for LLMs as well...
> > >
> > > On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> > >
> > > > Hi,
> > > >
> > > > Another option is to distribute MCP server publishing and SKILL/Agents
> > > > marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
> > > >
> > > > MCP servers are often (more often than not) not actual remote servers that
> > > > need to run "somewhere else". Many run locally, outside the agent's
> > > > sandbox, and expose the MCP server on a local stdin/connection, shielding
> > > > the MCP logic from the agent. Their server functionality is merely the API
> > > > they expose to local agents.
> > > >
> > > > I think there are two broad models we can discuss: centralized
> > > > (framework-based) and distributed (standards-based). Both can coexist
> > > > within the ASF. I can very easily **everything** we dp routed by ASF
> > > > `llmao` and "tight controlled" by the infrastructure / tooling - but also I
> > > > can very easily imagine that ASF contributors and maintainers use their own
> > > > LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
> > > > IMHO more scalable and more "Apache Way" approach. Both models have pros
> > > > and cons, but with the standards published in recent months, the barrier to
> > > > using the distributed approach has dramatically lowered.
> > > >
> > > > I personally strongly favor a distributed approach over a centralized one
> > > > for MCP servers, SKILLs, and Agent definitions. This allows each component
> > > > to be maintained and published separately by different PMCs directly in
> > > > their GitHub repos. It is sufficient for their code to be published
> > > > somewhere for the MCP server to be installed and run locally. For example,
> > > > the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
> > > > use in Magpie—are both simply installed and updated from the Comdev GitHub
> > > > repo.
> > > >
> > > > Standards for all of this were established in recent months (MCP 2.0 and
> > > > Agents Marketplace standards established in July/August).
> > > >
> > > > We recently converted Magpie to be installed from a distributed marketplace
> > > > in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
> > > > I am very happy with the test results; the open standards work smoothly
> > > > across many agent harnesses and are already quite powerful.
> > > >
> > > > For example, I used (rather than copying as I was doing so far) the threat
> > > > model skills from Alpha-Omega to handle threat modeling for our scan
> > > > preparation. I plan to make this accessible for all PMCs and non-ASF
> > > > projects—more details will be shared in Glasgow, where we plan to run
> > > > workshops addressing PMC needs around managing security issue volume
> > > > (including those from our scans).
> > > >
> > > > Moving forward, I plan to pursue what we previously discussed on the Magpie
> > > > devlist with Comdev and Incubator PMCs: distributing skill families back to
> > > > Comdev and Incubator so they can be managed and exposed directly by those
> > > > PMCs via distributed marketplaces, rather than hosting them in the Magpie
> > > > repository. There are few benefits to centralizing them in one place; it's
> > > > much better to have them managed by the respective PMCs and simply "link"
> > > > to them.
> > > >
> > > > Similarly, Legal could distribute their licensing, MCP, and SKILL
> > > > marketplace - and we could link to it from Magpie and expose the
> > > > "licencing" family of skills.
> > > >
> > > > This approach - I think - supports better scaling and distributes
> > > > governance in the true "Apache Way," giving each PMC autonomy over
> > > > implementation languages, scope, and publishing methods using open
> > > > standards.
> > > >
> > > > Best,
> > > > Jarek
> > > >
> > > > On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
> > > > wrote:
> > > >
> > > > > That's a valid point: magpie is definitely worth a try on podlings.
> > > > >
> > > > > Regards
> > > > > JB
> > > > >
> > > > > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> > > > > wrote:
> > > > > >
> > > > > > HI,
> > > > > >
> > > > > > On the AI side, Apache Magpie has a release skill that will find issues
> > > > > in release candidates for you. I suggest that prodlings try this out.
> > > > Also,
> > > > > the Policy MCP can be used to check releases. Having teh Policy MCP make
> > > > it
> > > > > consume a lot fewer tokens, faster, and up to date if the policy does
> > > > > change.
> > > > > >
> > > > > > Kind Regards,
> > > > > > Justin
> > > > > > ---------------------------------------------------------------------
> > > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > > >
> > > > >
> > > > > ---------------------------------------------------------------------
> > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > >
> > > > >
> > > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > For additional commands, e-mail: general-help@incubator.apache.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Edward Capriolo <ed...@gmail.com>.
On Thursday, September 10, 2026, Justin Mclean <ju...@classsoftware.com>
wrote:

> Hi,
>
> I agree with most of this. Writing down what we learn and putting it into
> shared tooling is what the Policy MCP and the Magpie release skill are for,
> and Paul's examples show it works. If a miss comes from a gap in the tools
> or docs, we should fix it once for everyone.
>
> Where I disagree is "if you can articulate it, it can be automated". The
> rule is easy to articulate: know where everything in the release came from
> and under what licence. Applying it needs information that isn't in the
> release.
>
> - A bundled dependency declares a licence, but that declaration can be
> wrong, dual-licensed as OR or AND, changed between versions, or only cover
> part of what's bundled. Paul's jQuery Treeview example is one of these.
>
> - What NOTICE needs depends on what the upstream NOTICE files say and
> what's actually been copied, not on whether a NOTICE file exists.
>
> - A file copied from a third-party project with an ASF header added passes
> RAT and every header check. Nothing in the file tells you it came from
> somewhere else. You find it because someone recognises it, searches for it,
> or asks where it came from.
>
> This is also already written down. Every release vote on this list records
> what was checked and what was found, and we've analysed about 1,600 of
> those vote threads from 2015 to 2025 [1]. Checklists and automation removed
> mechanical problems like bundled binaries, bad signatures, and checksum
> mismatches. What's left in most -1 votes now are legal and provenance
> issues, each in a different file in a different project, and those still
> need a person. Tools find more of them than they used to, and I'm all for
> that, but they usually flag things for someone to look at - they don't give
> a pass.
>
> Kind regards,
> Justin
>
> 1. https://cwiki.apache.org/confluence/spaces/INCUBATOR/
> pages/393677685/Release+Vote+Insights



The rat plugon already exists. Plugins can do just about anything and you
can write your own. If you give the English of what the verification
process is to an agent it could generate ultra-rat plugin in minutes  then
the obly requirements is project MUST use plugin x, avoiding should do work.


-- 
Sorry this was sent from mobile. Will do less grammar and spell check than
usual.

Re: [DISCUSS] Better License Verification Method

Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,

I agree with most of this. Writing down what we learn and putting it into shared tooling is what the Policy MCP and the Magpie release skill are for, and Paul's examples show it works. If a miss comes from a gap in the tools or docs, we should fix it once for everyone.

Where I disagree is "if you can articulate it, it can be automated". The rule is easy to articulate: know where everything in the release came from and under what licence. Applying it needs information that isn't in the release.

- A bundled dependency declares a licence, but that declaration can be wrong, dual-licensed as OR or AND, changed between versions, or only cover part of what's bundled. Paul's jQuery Treeview example is one of these.

- What NOTICE needs depends on what the upstream NOTICE files say and what's actually been copied, not on whether a NOTICE file exists.

- A file copied from a third-party project with an ASF header added passes RAT and every header check. Nothing in the file tells you it came from somewhere else. You find it because someone recognises it, searches for it, or asks where it came from.

This is also already written down. Every release vote on this list records what was checked and what was found, and we've analysed about 1,600 of those vote threads from 2015 to 2025 [1]. Checklists and automation removed mechanical problems like bundled binaries, bad signatures, and checksum mismatches. What's left in most -1 votes now are legal and provenance issues, each in a different file in a different project, and those still need a person. Tools find more of them than they used to, and I'm all for that, but they usually flag things for someone to look at - they don't give a pass.

Kind regards,
Justin

1. https://cwiki.apache.org/confluence/spaces/INCUBATOR/pages/393677685/Release+Vote+Insights

Re: [DISCUSS] Better License Verification Method

Posted by Stefan Krawczyk <st...@gmail.com>.
>  They don't tell you whether a bundled dependency's license is
compatible, whether NOTICE says the right thing, or whether a file with a
correct-looking header is genuinely under that license. A release can pass
all of them and still ship GPL-licensed code, this has happened more than
once.

I want to challenge that. If you can articulate it in instructions to
someone as to what to look for, it can be automated IMO. This is precisely
the esoteric knowledge that is best codified ideally in deterministic
tooling, or reusable documentation (e.g. skills or MCP) that LLMs can help
interpret for a human. I don't claim that this will be fool proof, but
missings things should just happen once, and then the collective tooling
and documentation updated to prevent it happening again - and the entire
Apache community benefitting - not just a single project. I think this
would make the onboarding experience of a podling a lot clearer too since
there would be a clearwr bar to clear.


On Thu, Sep 10, 2026, 2:18 PM Paul King <pa...@asert.com.au> wrote:

> For completeness, it also added:
> Remaining soft items, unchanged: the SDK zip is 116.5 MB against
> Infra's 100 MB recommendation, and jackson-core's shaded
> FastDoubleParser and Schubfach stay the deliberate exception.
>
> On Fri, Sep 11, 2026 at 7:16 AM Paul King <pa...@asert.com.au> wrote:
> >
> > Another example below of GroovyPolicyMCP used by Claude. As other's
> > have said, it doesn't replace further human and adhoc analysis.
> >
> > Cheers, Paul.
> >
> > ----
> >
> > GroovyPolicyMCP analysis:
> >
> > **Verdict: 6.0.0-RC-2 is clean. I would vote +1.** Every check on the
> > release checklist passed including signatures.
> >
> > **Tag and source**
> >
> > - The annotated tag on gitbox resolves to commit
> > 760601d45f652b90950f5dd3ba4e60f201691e11, matching the vote email, and
> > its parent is on origin/master.
> > - The source zip is identical to `git archive` at the tag. The only
> > absent files are the deliberate omissions: benchmark, the perf
> > dashboard, wrapper, governance docs, CI configs, and the ABI surface.
> > - No compiled code or archives in the source zip. RAT passes at the
> > tag. Version properties carry no SNAPSHOT.
> > - The JIRA changelog lists 44 unique issues, as the email says.
> >
> > **Signatures and checksums**
> >
> > | Location                           | Artifacts | Signed by
> > 6A65176A0FB1CD0B | Checksums                           |
> > | ---------------------------------- | --------- |
> > -------------------------- | ----------------------------------- |
> > | dist.apache.org zips               | 4         | 4 good
> >        | SHA-256 match                       |
> > | Nexus staging orgapachegroovy-1122 | 297       | 297 good
> >        | md5, sha1, sha256, sha512 all match |
> >
> > **LICENSE and NOTICE**
> >
> > - Binary and SDK zips: 18 of 18 license pointers resolve, no
> > unreferenced texts, NOTICE dated 2003-2026 with all the upstream
> > credits. All 88 bundled third-party jars covered, the same set as
> > today's snapshot.
> > - Docs zip: 4 of 4 pointers resolve.
> > - All 167 staged jars carry META-INF LICENSE and NOTICE with every
> > pointer resolving. Raw, shaded, grooid, sources and groovy-all-sources
> > declarations match their content. Sources, javadoc and groovydoc jars
> > contain no classes.
> > - The core, grooid and groovy-test-grooid jars inside the binary zip
> > are byte-identical to the staged ones, as are all 39 module jars in
> > lib/.
> > - All 43 poms declare the license, none reference non-Groovy SNAPSHOT
> > dependencies.
> >
> > **SBOMs** show 729 of 729 components licensed.
> >
> > On Wed, Sep 9, 2026 at 4:18 PM Paul King <pa...@asert.com.au> wrote:
> > >
> > > On Wed, Sep 9, 2026 at 3:30 PM tison <wa...@gmail.com> wrote:
> > > >
> > > > > To me this is all automateble. Why can't ASF have a blessed &
> centrally
> > > > > maintained way to check different file types at least as it
> pertains to
> > > > > licenses?
> > > >
> > > > It would be more than generous if you could give some concrete ideas
> > > > or a (referred) implementation.
> > >
> > > I find the various Gradle license plugins combined with a PolicyMCP
> > > goes a long way. Justin did various versions of a Python PolicyMCP
> > > implementation, I mostly use GroovyPolicyMCP (I have both normal
> > > Groovy and native versions, the latter faster than I can get combining
> > > Python/Rust).
> > >
> > > This is the kind of detail it gave on a recent Ivy release vote:
> > >
> > > **The jQuery Treeview plugin isn't listed in LICENSE.** Both zips
> > > bundle `jquery.treeview.js` (`asciidoc/js/` in src, `doc/js/` in bin).
> > > It's Treeview 1.2, copyright Jörn Zaefferer and Myles Angell,
> > > dual-licensed MIT/GPL — a separate work from jQuery core. The
> > > `LICENSE` file has a section for jQuery (John Resig) but nothing for
> > > Treeview, so it doesn't fully account for the package's exact content.
> > > The fix is a short "License for jQuery Treeview plugin" section taking
> > > the MIT option.
> > >
> > > One example from a recent Commons Collections release:
> > >
> > > 2. commons-collections4-4.6.0-javadoc.jar has no META-INF/LICENSE.txt
> or
> > >    NOTICE.txt, unlike the other jars. It carries the JDK's legal/
> directory
> > >    instead, where legal/LICENSE is the GPLv2 text (with Classpath
> Exception
> > >    via ASSEMBLY_EXCEPTION). Relatedly, the binary distribution bundles
> > >    jQuery 3.7.1 and jQuery UI (MIT) under apidocs/script-dir/, which
> the
> > >    root LICENSE.txt does not mention, though apidocs/legal/jquery.md
> and
> > >    jqueryUI.md are present. This is inherited from stock javadoc and is
> > >    common across ASF Java projects; the source release is unaffected.
> > >
> > > These happen to be using Claude but you can use your favourite agent.
> > >
> > > Paul.
> > >
> > > > Best,
> > > > tison.
> > > >
> > > > Stefan Krawczyk <st...@gmail.com> 于2026年9月9日周三 13:21写道:
> > > > >
> > > > > To me this is all automateble. Why can't ASF have a blessed &
> centrally
> > > > > maintained way to check different file types at least as it
> pertains to
> > > > > licenses? I think the majority of this can be 100% deterministic
> code
> > > > > without the need for LLMs as well...
> > > > >
> > > > > On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com>
> wrote:
> > > > >
> > > > > > Hi,
> > > > > >
> > > > > > Another option is to distribute MCP server publishing and
> SKILL/Agents
> > > > > > marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
> > > > > >
> > > > > > MCP servers are often (more often than not) not actual remote
> servers that
> > > > > > need to run "somewhere else". Many run locally, outside the
> agent's
> > > > > > sandbox, and expose the MCP server on a local stdin/connection,
> shielding
> > > > > > the MCP logic from the agent. Their server functionality is
> merely the API
> > > > > > they expose to local agents.
> > > > > >
> > > > > > I think there are two broad models we can discuss: centralized
> > > > > > (framework-based) and distributed (standards-based). Both can
> coexist
> > > > > > within the ASF. I can very easily **everything** we dp routed by
> ASF
> > > > > > `llmao` and "tight controlled" by the infrastructure / tooling -
> but also I
> > > > > > can very easily imagine that ASF contributors and maintainers
> use their own
> > > > > > LLM /  Subscriptions and local servers, connecting to ASF APIs -
> this is
> > > > > > IMHO more scalable and more "Apache Way" approach. Both models
> have pros
> > > > > > and cons, but with the standards published in recent months, the
> barrier to
> > > > > > using the distributed approach has dramatically lowered.
> > > > > >
> > > > > > I personally strongly favor a distributed approach over a
> centralized one
> > > > > > for MCP servers, SKILLs, and Agent definitions. This allows each
> component
> > > > > > to be maintained and published separately by different PMCs
> directly in
> > > > > > their GitHub repos. It is sufficient for their code to be
> published
> > > > > > somewhere for the MCP server to be installed and run locally.
> For example,
> > > > > > the MCP servers for Ponymail and Apache Project MCPs in Comdev
> —which we
> > > > > > use in Magpie—are both simply installed and updated from the
> Comdev GitHub
> > > > > > repo.
> > > > > >
> > > > > > Standards for all of this were established in recent months (MCP
> 2.0 and
> > > > > > Agents Marketplace standards established in July/August).
> > > > > >
> > > > > > We recently converted Magpie to be installed from a distributed
> marketplace
> > > > > > in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release.
> ) release.
> > > > > > I am very happy with the test results; the open standards work
> smoothly
> > > > > > across many agent harnesses and are already quite powerful.
> > > > > >
> > > > > > For example, I used (rather than copying as I was doing so far)
> the threat
> > > > > > model skills from Alpha-Omega to handle threat modeling for our
> scan
> > > > > > preparation. I plan to make this accessible for all PMCs and
> non-ASF
> > > > > > projects—more details will be shared in Glasgow, where we plan
> to run
> > > > > > workshops addressing PMC needs around managing security issue
> volume
> > > > > > (including those from our scans).
> > > > > >
> > > > > > Moving forward, I plan to pursue what we previously discussed on
> the Magpie
> > > > > > devlist with Comdev and Incubator PMCs: distributing skill
> families back to
> > > > > > Comdev and Incubator so they can be managed and exposed directly
> by those
> > > > > > PMCs via distributed marketplaces, rather than hosting them in
> the Magpie
> > > > > > repository. There are few benefits to centralizing them in one
> place; it's
> > > > > > much better to have them managed by the respective PMCs and
> simply "link"
> > > > > > to them.
> > > > > >
> > > > > > Similarly, Legal could distribute their licensing, MCP, and SKILL
> > > > > > marketplace - and we could link to it from Magpie and expose the
> > > > > > "licencing" family of skills.
> > > > > >
> > > > > > This approach - I think - supports better scaling and distributes
> > > > > > governance in the true "Apache Way," giving each PMC autonomy
> over
> > > > > > implementation languages, scope, and publishing methods using
> open
> > > > > > standards.
> > > > > >
> > > > > > Best,
> > > > > > Jarek
> > > > > >
> > > > > > On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <
> jb@nanthrax.net>
> > > > > > wrote:
> > > > > >
> > > > > > > That's a valid point: magpie is definitely worth a try on
> podlings.
> > > > > > >
> > > > > > > Regards
> > > > > > > JB
> > > > > > >
> > > > > > > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <
> justin@classsoftware.com>
> > > > > > > wrote:
> > > > > > > >
> > > > > > > > HI,
> > > > > > > >
> > > > > > > > On the AI side, Apache Magpie has a release skill that will
> find issues
> > > > > > > in release candidates for you. I suggest that prodlings try
> this out.
> > > > > > Also,
> > > > > > > the Policy MCP can be used to check releases. Having teh
> Policy MCP make
> > > > > > it
> > > > > > > consume a lot fewer tokens, faster, and up to date if the
> policy does
> > > > > > > change.
> > > > > > > >
> > > > > > > > Kind Regards,
> > > > > > > > Justin
> > > > > > > >
> ---------------------------------------------------------------------
> > > > > > > > To unsubscribe, e-mail:
> general-unsubscribe@incubator.apache.org
> > > > > > > > For additional commands, e-mail:
> general-help@incubator.apache.org
> > > > > > > >
> > > > > > >
> > > > > > >
> ---------------------------------------------------------------------
> > > > > > > To unsubscribe, e-mail:
> general-unsubscribe@incubator.apache.org
> > > > > > > For additional commands, e-mail:
> general-help@incubator.apache.org
> > > > > > >
> > > > > > >
> > > > > >
> > > >
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>
>

Re: [DISCUSS] Better License Verification Method

Posted by Paul King <pa...@asert.com.au>.
For completeness, it also added:
Remaining soft items, unchanged: the SDK zip is 116.5 MB against
Infra's 100 MB recommendation, and jackson-core's shaded
FastDoubleParser and Schubfach stay the deliberate exception.

On Fri, Sep 11, 2026 at 7:16 AM Paul King <pa...@asert.com.au> wrote:
>
> Another example below of GroovyPolicyMCP used by Claude. As other's
> have said, it doesn't replace further human and adhoc analysis.
>
> Cheers, Paul.
>
> ----
>
> GroovyPolicyMCP analysis:
>
> **Verdict: 6.0.0-RC-2 is clean. I would vote +1.** Every check on the
> release checklist passed including signatures.
>
> **Tag and source**
>
> - The annotated tag on gitbox resolves to commit
> 760601d45f652b90950f5dd3ba4e60f201691e11, matching the vote email, and
> its parent is on origin/master.
> - The source zip is identical to `git archive` at the tag. The only
> absent files are the deliberate omissions: benchmark, the perf
> dashboard, wrapper, governance docs, CI configs, and the ABI surface.
> - No compiled code or archives in the source zip. RAT passes at the
> tag. Version properties carry no SNAPSHOT.
> - The JIRA changelog lists 44 unique issues, as the email says.
>
> **Signatures and checksums**
>
> | Location                           | Artifacts | Signed by
> 6A65176A0FB1CD0B | Checksums                           |
> | ---------------------------------- | --------- |
> -------------------------- | ----------------------------------- |
> | dist.apache.org zips               | 4         | 4 good
>        | SHA-256 match                       |
> | Nexus staging orgapachegroovy-1122 | 297       | 297 good
>        | md5, sha1, sha256, sha512 all match |
>
> **LICENSE and NOTICE**
>
> - Binary and SDK zips: 18 of 18 license pointers resolve, no
> unreferenced texts, NOTICE dated 2003-2026 with all the upstream
> credits. All 88 bundled third-party jars covered, the same set as
> today's snapshot.
> - Docs zip: 4 of 4 pointers resolve.
> - All 167 staged jars carry META-INF LICENSE and NOTICE with every
> pointer resolving. Raw, shaded, grooid, sources and groovy-all-sources
> declarations match their content. Sources, javadoc and groovydoc jars
> contain no classes.
> - The core, grooid and groovy-test-grooid jars inside the binary zip
> are byte-identical to the staged ones, as are all 39 module jars in
> lib/.
> - All 43 poms declare the license, none reference non-Groovy SNAPSHOT
> dependencies.
>
> **SBOMs** show 729 of 729 components licensed.
>
> On Wed, Sep 9, 2026 at 4:18 PM Paul King <pa...@asert.com.au> wrote:
> >
> > On Wed, Sep 9, 2026 at 3:30 PM tison <wa...@gmail.com> wrote:
> > >
> > > > To me this is all automateble. Why can't ASF have a blessed & centrally
> > > > maintained way to check different file types at least as it pertains to
> > > > licenses?
> > >
> > > It would be more than generous if you could give some concrete ideas
> > > or a (referred) implementation.
> >
> > I find the various Gradle license plugins combined with a PolicyMCP
> > goes a long way. Justin did various versions of a Python PolicyMCP
> > implementation, I mostly use GroovyPolicyMCP (I have both normal
> > Groovy and native versions, the latter faster than I can get combining
> > Python/Rust).
> >
> > This is the kind of detail it gave on a recent Ivy release vote:
> >
> > **The jQuery Treeview plugin isn't listed in LICENSE.** Both zips
> > bundle `jquery.treeview.js` (`asciidoc/js/` in src, `doc/js/` in bin).
> > It's Treeview 1.2, copyright Jörn Zaefferer and Myles Angell,
> > dual-licensed MIT/GPL — a separate work from jQuery core. The
> > `LICENSE` file has a section for jQuery (John Resig) but nothing for
> > Treeview, so it doesn't fully account for the package's exact content.
> > The fix is a short "License for jQuery Treeview plugin" section taking
> > the MIT option.
> >
> > One example from a recent Commons Collections release:
> >
> > 2. commons-collections4-4.6.0-javadoc.jar has no META-INF/LICENSE.txt or
> >    NOTICE.txt, unlike the other jars. It carries the JDK's legal/ directory
> >    instead, where legal/LICENSE is the GPLv2 text (with Classpath Exception
> >    via ASSEMBLY_EXCEPTION). Relatedly, the binary distribution bundles
> >    jQuery 3.7.1 and jQuery UI (MIT) under apidocs/script-dir/, which the
> >    root LICENSE.txt does not mention, though apidocs/legal/jquery.md and
> >    jqueryUI.md are present. This is inherited from stock javadoc and is
> >    common across ASF Java projects; the source release is unaffected.
> >
> > These happen to be using Claude but you can use your favourite agent.
> >
> > Paul.
> >
> > > Best,
> > > tison.
> > >
> > > Stefan Krawczyk <st...@gmail.com> 于2026年9月9日周三 13:21写道:
> > > >
> > > > To me this is all automateble. Why can't ASF have a blessed & centrally
> > > > maintained way to check different file types at least as it pertains to
> > > > licenses? I think the majority of this can be 100% deterministic code
> > > > without the need for LLMs as well...
> > > >
> > > > On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> > > >
> > > > > Hi,
> > > > >
> > > > > Another option is to distribute MCP server publishing and SKILL/Agents
> > > > > marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
> > > > >
> > > > > MCP servers are often (more often than not) not actual remote servers that
> > > > > need to run "somewhere else". Many run locally, outside the agent's
> > > > > sandbox, and expose the MCP server on a local stdin/connection, shielding
> > > > > the MCP logic from the agent. Their server functionality is merely the API
> > > > > they expose to local agents.
> > > > >
> > > > > I think there are two broad models we can discuss: centralized
> > > > > (framework-based) and distributed (standards-based). Both can coexist
> > > > > within the ASF. I can very easily **everything** we dp routed by ASF
> > > > > `llmao` and "tight controlled" by the infrastructure / tooling - but also I
> > > > > can very easily imagine that ASF contributors and maintainers use their own
> > > > > LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
> > > > > IMHO more scalable and more "Apache Way" approach. Both models have pros
> > > > > and cons, but with the standards published in recent months, the barrier to
> > > > > using the distributed approach has dramatically lowered.
> > > > >
> > > > > I personally strongly favor a distributed approach over a centralized one
> > > > > for MCP servers, SKILLs, and Agent definitions. This allows each component
> > > > > to be maintained and published separately by different PMCs directly in
> > > > > their GitHub repos. It is sufficient for their code to be published
> > > > > somewhere for the MCP server to be installed and run locally. For example,
> > > > > the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
> > > > > use in Magpie—are both simply installed and updated from the Comdev GitHub
> > > > > repo.
> > > > >
> > > > > Standards for all of this were established in recent months (MCP 2.0 and
> > > > > Agents Marketplace standards established in July/August).
> > > > >
> > > > > We recently converted Magpie to be installed from a distributed marketplace
> > > > > in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
> > > > > I am very happy with the test results; the open standards work smoothly
> > > > > across many agent harnesses and are already quite powerful.
> > > > >
> > > > > For example, I used (rather than copying as I was doing so far) the threat
> > > > > model skills from Alpha-Omega to handle threat modeling for our scan
> > > > > preparation. I plan to make this accessible for all PMCs and non-ASF
> > > > > projects—more details will be shared in Glasgow, where we plan to run
> > > > > workshops addressing PMC needs around managing security issue volume
> > > > > (including those from our scans).
> > > > >
> > > > > Moving forward, I plan to pursue what we previously discussed on the Magpie
> > > > > devlist with Comdev and Incubator PMCs: distributing skill families back to
> > > > > Comdev and Incubator so they can be managed and exposed directly by those
> > > > > PMCs via distributed marketplaces, rather than hosting them in the Magpie
> > > > > repository. There are few benefits to centralizing them in one place; it's
> > > > > much better to have them managed by the respective PMCs and simply "link"
> > > > > to them.
> > > > >
> > > > > Similarly, Legal could distribute their licensing, MCP, and SKILL
> > > > > marketplace - and we could link to it from Magpie and expose the
> > > > > "licencing" family of skills.
> > > > >
> > > > > This approach - I think - supports better scaling and distributes
> > > > > governance in the true "Apache Way," giving each PMC autonomy over
> > > > > implementation languages, scope, and publishing methods using open
> > > > > standards.
> > > > >
> > > > > Best,
> > > > > Jarek
> > > > >
> > > > > On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
> > > > > wrote:
> > > > >
> > > > > > That's a valid point: magpie is definitely worth a try on podlings.
> > > > > >
> > > > > > Regards
> > > > > > JB
> > > > > >
> > > > > > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> > > > > > wrote:
> > > > > > >
> > > > > > > HI,
> > > > > > >
> > > > > > > On the AI side, Apache Magpie has a release skill that will find issues
> > > > > > in release candidates for you. I suggest that prodlings try this out.
> > > > > Also,
> > > > > > the Policy MCP can be used to check releases. Having teh Policy MCP make
> > > > > it
> > > > > > consume a lot fewer tokens, faster, and up to date if the policy does
> > > > > > change.
> > > > > > >
> > > > > > > Kind Regards,
> > > > > > > Justin
> > > > > > > ---------------------------------------------------------------------
> > > > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > > > >
> > > > > >
> > > > > > ---------------------------------------------------------------------
> > > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > > >
> > > > > >
> > > > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > For additional commands, e-mail: general-help@incubator.apache.org
> > >

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Paul King <pa...@asert.com.au>.
Another example below of GroovyPolicyMCP used by Claude. As other's
have said, it doesn't replace further human and adhoc analysis.

Cheers, Paul.

----

GroovyPolicyMCP analysis:

**Verdict: 6.0.0-RC-2 is clean. I would vote +1.** Every check on the
release checklist passed including signatures.

**Tag and source**

- The annotated tag on gitbox resolves to commit
760601d45f652b90950f5dd3ba4e60f201691e11, matching the vote email, and
its parent is on origin/master.
- The source zip is identical to `git archive` at the tag. The only
absent files are the deliberate omissions: benchmark, the perf
dashboard, wrapper, governance docs, CI configs, and the ABI surface.
- No compiled code or archives in the source zip. RAT passes at the
tag. Version properties carry no SNAPSHOT.
- The JIRA changelog lists 44 unique issues, as the email says.

**Signatures and checksums**

| Location                           | Artifacts | Signed by
6A65176A0FB1CD0B | Checksums                           |
| ---------------------------------- | --------- |
-------------------------- | ----------------------------------- |
| dist.apache.org zips               | 4         | 4 good
       | SHA-256 match                       |
| Nexus staging orgapachegroovy-1122 | 297       | 297 good
       | md5, sha1, sha256, sha512 all match |

**LICENSE and NOTICE**

- Binary and SDK zips: 18 of 18 license pointers resolve, no
unreferenced texts, NOTICE dated 2003-2026 with all the upstream
credits. All 88 bundled third-party jars covered, the same set as
today's snapshot.
- Docs zip: 4 of 4 pointers resolve.
- All 167 staged jars carry META-INF LICENSE and NOTICE with every
pointer resolving. Raw, shaded, grooid, sources and groovy-all-sources
declarations match their content. Sources, javadoc and groovydoc jars
contain no classes.
- The core, grooid and groovy-test-grooid jars inside the binary zip
are byte-identical to the staged ones, as are all 39 module jars in
lib/.
- All 43 poms declare the license, none reference non-Groovy SNAPSHOT
dependencies.

**SBOMs** show 729 of 729 components licensed.

On Wed, Sep 9, 2026 at 4:18 PM Paul King <pa...@asert.com.au> wrote:
>
> On Wed, Sep 9, 2026 at 3:30 PM tison <wa...@gmail.com> wrote:
> >
> > > To me this is all automateble. Why can't ASF have a blessed & centrally
> > > maintained way to check different file types at least as it pertains to
> > > licenses?
> >
> > It would be more than generous if you could give some concrete ideas
> > or a (referred) implementation.
>
> I find the various Gradle license plugins combined with a PolicyMCP
> goes a long way. Justin did various versions of a Python PolicyMCP
> implementation, I mostly use GroovyPolicyMCP (I have both normal
> Groovy and native versions, the latter faster than I can get combining
> Python/Rust).
>
> This is the kind of detail it gave on a recent Ivy release vote:
>
> **The jQuery Treeview plugin isn't listed in LICENSE.** Both zips
> bundle `jquery.treeview.js` (`asciidoc/js/` in src, `doc/js/` in bin).
> It's Treeview 1.2, copyright Jörn Zaefferer and Myles Angell,
> dual-licensed MIT/GPL — a separate work from jQuery core. The
> `LICENSE` file has a section for jQuery (John Resig) but nothing for
> Treeview, so it doesn't fully account for the package's exact content.
> The fix is a short "License for jQuery Treeview plugin" section taking
> the MIT option.
>
> One example from a recent Commons Collections release:
>
> 2. commons-collections4-4.6.0-javadoc.jar has no META-INF/LICENSE.txt or
>    NOTICE.txt, unlike the other jars. It carries the JDK's legal/ directory
>    instead, where legal/LICENSE is the GPLv2 text (with Classpath Exception
>    via ASSEMBLY_EXCEPTION). Relatedly, the binary distribution bundles
>    jQuery 3.7.1 and jQuery UI (MIT) under apidocs/script-dir/, which the
>    root LICENSE.txt does not mention, though apidocs/legal/jquery.md and
>    jqueryUI.md are present. This is inherited from stock javadoc and is
>    common across ASF Java projects; the source release is unaffected.
>
> These happen to be using Claude but you can use your favourite agent.
>
> Paul.
>
> > Best,
> > tison.
> >
> > Stefan Krawczyk <st...@gmail.com> 于2026年9月9日周三 13:21写道:
> > >
> > > To me this is all automateble. Why can't ASF have a blessed & centrally
> > > maintained way to check different file types at least as it pertains to
> > > licenses? I think the majority of this can be 100% deterministic code
> > > without the need for LLMs as well...
> > >
> > > On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> > >
> > > > Hi,
> > > >
> > > > Another option is to distribute MCP server publishing and SKILL/Agents
> > > > marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
> > > >
> > > > MCP servers are often (more often than not) not actual remote servers that
> > > > need to run "somewhere else". Many run locally, outside the agent's
> > > > sandbox, and expose the MCP server on a local stdin/connection, shielding
> > > > the MCP logic from the agent. Their server functionality is merely the API
> > > > they expose to local agents.
> > > >
> > > > I think there are two broad models we can discuss: centralized
> > > > (framework-based) and distributed (standards-based). Both can coexist
> > > > within the ASF. I can very easily **everything** we dp routed by ASF
> > > > `llmao` and "tight controlled" by the infrastructure / tooling - but also I
> > > > can very easily imagine that ASF contributors and maintainers use their own
> > > > LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
> > > > IMHO more scalable and more "Apache Way" approach. Both models have pros
> > > > and cons, but with the standards published in recent months, the barrier to
> > > > using the distributed approach has dramatically lowered.
> > > >
> > > > I personally strongly favor a distributed approach over a centralized one
> > > > for MCP servers, SKILLs, and Agent definitions. This allows each component
> > > > to be maintained and published separately by different PMCs directly in
> > > > their GitHub repos. It is sufficient for their code to be published
> > > > somewhere for the MCP server to be installed and run locally. For example,
> > > > the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
> > > > use in Magpie—are both simply installed and updated from the Comdev GitHub
> > > > repo.
> > > >
> > > > Standards for all of this were established in recent months (MCP 2.0 and
> > > > Agents Marketplace standards established in July/August).
> > > >
> > > > We recently converted Magpie to be installed from a distributed marketplace
> > > > in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
> > > > I am very happy with the test results; the open standards work smoothly
> > > > across many agent harnesses and are already quite powerful.
> > > >
> > > > For example, I used (rather than copying as I was doing so far) the threat
> > > > model skills from Alpha-Omega to handle threat modeling for our scan
> > > > preparation. I plan to make this accessible for all PMCs and non-ASF
> > > > projects—more details will be shared in Glasgow, where we plan to run
> > > > workshops addressing PMC needs around managing security issue volume
> > > > (including those from our scans).
> > > >
> > > > Moving forward, I plan to pursue what we previously discussed on the Magpie
> > > > devlist with Comdev and Incubator PMCs: distributing skill families back to
> > > > Comdev and Incubator so they can be managed and exposed directly by those
> > > > PMCs via distributed marketplaces, rather than hosting them in the Magpie
> > > > repository. There are few benefits to centralizing them in one place; it's
> > > > much better to have them managed by the respective PMCs and simply "link"
> > > > to them.
> > > >
> > > > Similarly, Legal could distribute their licensing, MCP, and SKILL
> > > > marketplace - and we could link to it from Magpie and expose the
> > > > "licencing" family of skills.
> > > >
> > > > This approach - I think - supports better scaling and distributes
> > > > governance in the true "Apache Way," giving each PMC autonomy over
> > > > implementation languages, scope, and publishing methods using open
> > > > standards.
> > > >
> > > > Best,
> > > > Jarek
> > > >
> > > > On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
> > > > wrote:
> > > >
> > > > > That's a valid point: magpie is definitely worth a try on podlings.
> > > > >
> > > > > Regards
> > > > > JB
> > > > >
> > > > > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> > > > > wrote:
> > > > > >
> > > > > > HI,
> > > > > >
> > > > > > On the AI side, Apache Magpie has a release skill that will find issues
> > > > > in release candidates for you. I suggest that prodlings try this out.
> > > > Also,
> > > > > the Policy MCP can be used to check releases. Having teh Policy MCP make
> > > > it
> > > > > consume a lot fewer tokens, faster, and up to date if the policy does
> > > > > change.
> > > > > >
> > > > > > Kind Regards,
> > > > > > Justin
> > > > > > ---------------------------------------------------------------------
> > > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > > >
> > > > >
> > > > > ---------------------------------------------------------------------
> > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > >
> > > > >
> > > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > For additional commands, e-mail: general-help@incubator.apache.org
> >

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Paul King <pa...@asert.com.au>.
On Wed, Sep 9, 2026 at 3:30 PM tison <wa...@gmail.com> wrote:
>
> > To me this is all automateble. Why can't ASF have a blessed & centrally
> > maintained way to check different file types at least as it pertains to
> > licenses?
>
> It would be more than generous if you could give some concrete ideas
> or a (referred) implementation.

I find the various Gradle license plugins combined with a PolicyMCP
goes a long way. Justin did various versions of a Python PolicyMCP
implementation, I mostly use GroovyPolicyMCP (I have both normal
Groovy and native versions, the latter faster than I can get combining
Python/Rust).

This is the kind of detail it gave on a recent Ivy release vote:

**The jQuery Treeview plugin isn't listed in LICENSE.** Both zips
bundle `jquery.treeview.js` (`asciidoc/js/` in src, `doc/js/` in bin).
It's Treeview 1.2, copyright Jörn Zaefferer and Myles Angell,
dual-licensed MIT/GPL — a separate work from jQuery core. The
`LICENSE` file has a section for jQuery (John Resig) but nothing for
Treeview, so it doesn't fully account for the package's exact content.
The fix is a short "License for jQuery Treeview plugin" section taking
the MIT option.

One example from a recent Commons Collections release:

2. commons-collections4-4.6.0-javadoc.jar has no META-INF/LICENSE.txt or
   NOTICE.txt, unlike the other jars. It carries the JDK's legal/ directory
   instead, where legal/LICENSE is the GPLv2 text (with Classpath Exception
   via ASSEMBLY_EXCEPTION). Relatedly, the binary distribution bundles
   jQuery 3.7.1 and jQuery UI (MIT) under apidocs/script-dir/, which the
   root LICENSE.txt does not mention, though apidocs/legal/jquery.md and
   jqueryUI.md are present. This is inherited from stock javadoc and is
   common across ASF Java projects; the source release is unaffected.

These happen to be using Claude but you can use your favourite agent.

Paul.

> Best,
> tison.
>
> Stefan Krawczyk <st...@gmail.com> 于2026年9月9日周三 13:21写道:
> >
> > To me this is all automateble. Why can't ASF have a blessed & centrally
> > maintained way to check different file types at least as it pertains to
> > licenses? I think the majority of this can be 100% deterministic code
> > without the need for LLMs as well...
> >
> > On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com> wrote:
> >
> > > Hi,
> > >
> > > Another option is to distribute MCP server publishing and SKILL/Agents
> > > marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
> > >
> > > MCP servers are often (more often than not) not actual remote servers that
> > > need to run "somewhere else". Many run locally, outside the agent's
> > > sandbox, and expose the MCP server on a local stdin/connection, shielding
> > > the MCP logic from the agent. Their server functionality is merely the API
> > > they expose to local agents.
> > >
> > > I think there are two broad models we can discuss: centralized
> > > (framework-based) and distributed (standards-based). Both can coexist
> > > within the ASF. I can very easily **everything** we dp routed by ASF
> > > `llmao` and "tight controlled" by the infrastructure / tooling - but also I
> > > can very easily imagine that ASF contributors and maintainers use their own
> > > LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
> > > IMHO more scalable and more "Apache Way" approach. Both models have pros
> > > and cons, but with the standards published in recent months, the barrier to
> > > using the distributed approach has dramatically lowered.
> > >
> > > I personally strongly favor a distributed approach over a centralized one
> > > for MCP servers, SKILLs, and Agent definitions. This allows each component
> > > to be maintained and published separately by different PMCs directly in
> > > their GitHub repos. It is sufficient for their code to be published
> > > somewhere for the MCP server to be installed and run locally. For example,
> > > the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
> > > use in Magpie—are both simply installed and updated from the Comdev GitHub
> > > repo.
> > >
> > > Standards for all of this were established in recent months (MCP 2.0 and
> > > Agents Marketplace standards established in July/August).
> > >
> > > We recently converted Magpie to be installed from a distributed marketplace
> > > in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
> > > I am very happy with the test results; the open standards work smoothly
> > > across many agent harnesses and are already quite powerful.
> > >
> > > For example, I used (rather than copying as I was doing so far) the threat
> > > model skills from Alpha-Omega to handle threat modeling for our scan
> > > preparation. I plan to make this accessible for all PMCs and non-ASF
> > > projects—more details will be shared in Glasgow, where we plan to run
> > > workshops addressing PMC needs around managing security issue volume
> > > (including those from our scans).
> > >
> > > Moving forward, I plan to pursue what we previously discussed on the Magpie
> > > devlist with Comdev and Incubator PMCs: distributing skill families back to
> > > Comdev and Incubator so they can be managed and exposed directly by those
> > > PMCs via distributed marketplaces, rather than hosting them in the Magpie
> > > repository. There are few benefits to centralizing them in one place; it's
> > > much better to have them managed by the respective PMCs and simply "link"
> > > to them.
> > >
> > > Similarly, Legal could distribute their licensing, MCP, and SKILL
> > > marketplace - and we could link to it from Magpie and expose the
> > > "licencing" family of skills.
> > >
> > > This approach - I think - supports better scaling and distributes
> > > governance in the true "Apache Way," giving each PMC autonomy over
> > > implementation languages, scope, and publishing methods using open
> > > standards.
> > >
> > > Best,
> > > Jarek
> > >
> > > On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
> > > wrote:
> > >
> > > > That's a valid point: magpie is definitely worth a try on podlings.
> > > >
> > > > Regards
> > > > JB
> > > >
> > > > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> > > > wrote:
> > > > >
> > > > > HI,
> > > > >
> > > > > On the AI side, Apache Magpie has a release skill that will find issues
> > > > in release candidates for you. I suggest that prodlings try this out.
> > > Also,
> > > > the Policy MCP can be used to check releases. Having teh Policy MCP make
> > > it
> > > > consume a lot fewer tokens, faster, and up to date if the policy does
> > > > change.
> > > > >
> > > > > Kind Regards,
> > > > > Justin
> > > > > ---------------------------------------------------------------------
> > > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > > >
> > > >
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > >
> > > >
> > >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by tison <wa...@gmail.com>.
> To me this is all automateble. Why can't ASF have a blessed & centrally
> maintained way to check different file types at least as it pertains to
> licenses?

It would be more than generous if you could give some concrete ideas
or a (referred) implementation.

Best,
tison.

Stefan Krawczyk <st...@gmail.com> 于2026年9月9日周三 13:21写道:
>
> To me this is all automateble. Why can't ASF have a blessed & centrally
> maintained way to check different file types at least as it pertains to
> licenses? I think the majority of this can be 100% deterministic code
> without the need for LLMs as well...
>
> On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com> wrote:
>
> > Hi,
> >
> > Another option is to distribute MCP server publishing and SKILL/Agents
> > marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
> >
> > MCP servers are often (more often than not) not actual remote servers that
> > need to run "somewhere else". Many run locally, outside the agent's
> > sandbox, and expose the MCP server on a local stdin/connection, shielding
> > the MCP logic from the agent. Their server functionality is merely the API
> > they expose to local agents.
> >
> > I think there are two broad models we can discuss: centralized
> > (framework-based) and distributed (standards-based). Both can coexist
> > within the ASF. I can very easily **everything** we dp routed by ASF
> > `llmao` and "tight controlled" by the infrastructure / tooling - but also I
> > can very easily imagine that ASF contributors and maintainers use their own
> > LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
> > IMHO more scalable and more "Apache Way" approach. Both models have pros
> > and cons, but with the standards published in recent months, the barrier to
> > using the distributed approach has dramatically lowered.
> >
> > I personally strongly favor a distributed approach over a centralized one
> > for MCP servers, SKILLs, and Agent definitions. This allows each component
> > to be maintained and published separately by different PMCs directly in
> > their GitHub repos. It is sufficient for their code to be published
> > somewhere for the MCP server to be installed and run locally. For example,
> > the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
> > use in Magpie—are both simply installed and updated from the Comdev GitHub
> > repo.
> >
> > Standards for all of this were established in recent months (MCP 2.0 and
> > Agents Marketplace standards established in July/August).
> >
> > We recently converted Magpie to be installed from a distributed marketplace
> > in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
> > I am very happy with the test results; the open standards work smoothly
> > across many agent harnesses and are already quite powerful.
> >
> > For example, I used (rather than copying as I was doing so far) the threat
> > model skills from Alpha-Omega to handle threat modeling for our scan
> > preparation. I plan to make this accessible for all PMCs and non-ASF
> > projects—more details will be shared in Glasgow, where we plan to run
> > workshops addressing PMC needs around managing security issue volume
> > (including those from our scans).
> >
> > Moving forward, I plan to pursue what we previously discussed on the Magpie
> > devlist with Comdev and Incubator PMCs: distributing skill families back to
> > Comdev and Incubator so they can be managed and exposed directly by those
> > PMCs via distributed marketplaces, rather than hosting them in the Magpie
> > repository. There are few benefits to centralizing them in one place; it's
> > much better to have them managed by the respective PMCs and simply "link"
> > to them.
> >
> > Similarly, Legal could distribute their licensing, MCP, and SKILL
> > marketplace - and we could link to it from Magpie and expose the
> > "licencing" family of skills.
> >
> > This approach - I think - supports better scaling and distributes
> > governance in the true "Apache Way," giving each PMC autonomy over
> > implementation languages, scope, and publishing methods using open
> > standards.
> >
> > Best,
> > Jarek
> >
> > On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
> > wrote:
> >
> > > That's a valid point: magpie is definitely worth a try on podlings.
> > >
> > > Regards
> > > JB
> > >
> > > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> > > wrote:
> > > >
> > > > HI,
> > > >
> > > > On the AI side, Apache Magpie has a release skill that will find issues
> > > in release candidates for you. I suggest that prodlings try this out.
> > Also,
> > > the Policy MCP can be used to check releases. Having teh Policy MCP make
> > it
> > > consume a lot fewer tokens, faster, and up to date if the policy does
> > > change.
> > > >
> > > > Kind Regards,
> > > > Justin
> > > > ---------------------------------------------------------------------
> > > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > > For additional commands, e-mail: general-help@incubator.apache.org
> > > >
> > >
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > For additional commands, e-mail: general-help@incubator.apache.org
> > >
> > >
> >

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Dave Fisher <wa...@apache.org>.
Hi -

A skill or agent that helps with assembling the LICENSE, NOTICE, and .rat-excludes file gives a good way to focus on what humans should check in the source code. I think that an agreed tool that uses an MCP on legal-discuss@ and general@ could really help projects do their diligence often - from a periodic workflow, but not from CI on every PR (we don’t have the budget for the number of tokens that would take).

Once such a skill and agent exists and is proven I would add this to ATR as one of the checks that is always done. It would be adjacent to the RAT check already done and would provide an important adjunct to the LICENSE and NOTICE checks which are not much more than existence checks.

SBOM tools and some of the databases the ecosystem is building are a good start, but again the license annotations and whether a dependency with multiple licenses is an OR or an AND needs a check. On top of that a dependency’s license could be changed. There are also additional risk based considerations with dependencies like is it maintained? In ATR’s presentation of SBOM information we are trying to highlight concerns and we acknowledge that some of these are bad data and missing mappings.

Best,
Dave

> On Sep 9, 2026, at 8:50 AM, Justin Mclean <ju...@classsoftware.com> wrote:
> 
> Hi,
> 
> Automation cannot solve this problem; it only solves part of it. We have RAT, and most projects run it as part of their build.
> 
> I've been reviewing incubator releases for well over a decade, and the issues I raise are not ones RAT would have caught. A recent candidate used HawkEye, the PPMC voted +1 and found no issues, and the IPMC voted -1 because of valid issues.
> 
> Header checks and dependency scans are good at what they cover. They don't tell you whether a bundled dependency's license is compatible, whether NOTICE says the right thing, or whether a file with a correct-looking header is genuinely under that license. A release can pass all of them and still ship GPL-licensed code, this has happened more than once.
> 
> Tools narrow down what a reviewer has to look at, but they don't remove the reviewer, and we cannot 100% automate this.
> 
> Kind regards,
> Justin
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Justin Mclean <ju...@classsoftware.com>.
Hi,

Automation cannot solve this problem; it only solves part of it. We have RAT, and most projects run it as part of their build.

I've been reviewing incubator releases for well over a decade, and the issues I raise are not ones RAT would have caught. A recent candidate used HawkEye, the PPMC voted +1 and found no issues, and the IPMC voted -1 because of valid issues.

Header checks and dependency scans are good at what they cover. They don't tell you whether a bundled dependency's license is compatible, whether NOTICE says the right thing, or whether a file with a correct-looking header is genuinely under that license. A release can pass all of them and still ship GPL-licensed code, this has happened more than once.

Tools narrow down what a reviewer has to look at, but they don't remove the reviewer, and we cannot 100% automate this.

Kind regards,
Justin
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Stefan Krawczyk <st...@gmail.com>.
To me this is all automateble. Why can't ASF have a blessed & centrally
maintained way to check different file types at least as it pertains to
licenses? I think the majority of this can be 100% deterministic code
without the need for LLMs as well...

On Tue, Sep 8, 2026, 2:52 AM Jarek Potiuk <ja...@potiuk.com> wrote:

> Hi,
>
> Another option is to distribute MCP server publishing and SKILL/Agents
> marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.
>
> MCP servers are often (more often than not) not actual remote servers that
> need to run "somewhere else". Many run locally, outside the agent's
> sandbox, and expose the MCP server on a local stdin/connection, shielding
> the MCP logic from the agent. Their server functionality is merely the API
> they expose to local agents.
>
> I think there are two broad models we can discuss: centralized
> (framework-based) and distributed (standards-based). Both can coexist
> within the ASF. I can very easily **everything** we dp routed by ASF
> `llmao` and "tight controlled" by the infrastructure / tooling - but also I
> can very easily imagine that ASF contributors and maintainers use their own
> LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
> IMHO more scalable and more "Apache Way" approach. Both models have pros
> and cons, but with the standards published in recent months, the barrier to
> using the distributed approach has dramatically lowered.
>
> I personally strongly favor a distributed approach over a centralized one
> for MCP servers, SKILLs, and Agent definitions. This allows each component
> to be maintained and published separately by different PMCs directly in
> their GitHub repos. It is sufficient for their code to be published
> somewhere for the MCP server to be installed and run locally. For example,
> the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
> use in Magpie—are both simply installed and updated from the Comdev GitHub
> repo.
>
> Standards for all of this were established in recent months (MCP 2.0 and
> Agents Marketplace standards established in July/August).
>
> We recently converted Magpie to be installed from a distributed marketplace
> in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
> I am very happy with the test results; the open standards work smoothly
> across many agent harnesses and are already quite powerful.
>
> For example, I used (rather than copying as I was doing so far) the threat
> model skills from Alpha-Omega to handle threat modeling for our scan
> preparation. I plan to make this accessible for all PMCs and non-ASF
> projects—more details will be shared in Glasgow, where we plan to run
> workshops addressing PMC needs around managing security issue volume
> (including those from our scans).
>
> Moving forward, I plan to pursue what we previously discussed on the Magpie
> devlist with Comdev and Incubator PMCs: distributing skill families back to
> Comdev and Incubator so they can be managed and exposed directly by those
> PMCs via distributed marketplaces, rather than hosting them in the Magpie
> repository. There are few benefits to centralizing them in one place; it's
> much better to have them managed by the respective PMCs and simply "link"
> to them.
>
> Similarly, Legal could distribute their licensing, MCP, and SKILL
> marketplace - and we could link to it from Magpie and expose the
> "licencing" family of skills.
>
> This approach - I think - supports better scaling and distributes
> governance in the true "Apache Way," giving each PMC autonomy over
> implementation languages, scope, and publishing methods using open
> standards.
>
> Best,
> Jarek
>
> On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
> wrote:
>
> > That's a valid point: magpie is definitely worth a try on podlings.
> >
> > Regards
> > JB
> >
> > On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> > wrote:
> > >
> > > HI,
> > >
> > > On the AI side, Apache Magpie has a release skill that will find issues
> > in release candidates for you. I suggest that prodlings try this out.
> Also,
> > the Policy MCP can be used to check releases. Having teh Policy MCP make
> it
> > consume a lot fewer tokens, faster, and up to date if the policy does
> > change.
> > >
> > > Kind Regards,
> > > Justin
> > > ---------------------------------------------------------------------
> > > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > > For additional commands, e-mail: general-help@incubator.apache.org
> > >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > For additional commands, e-mail: general-help@incubator.apache.org
> >
> >
>

Re: [DISCUSS] Better License Verification Method

Posted by Jarek Potiuk <ja...@potiuk.com>.
Hi,

Another option is to distribute MCP server publishing and SKILL/Agents
marketplaces. Or, as JBO mentioned, use CLIs and remote APIs.

MCP servers are often (more often than not) not actual remote servers that
need to run "somewhere else". Many run locally, outside the agent's
sandbox, and expose the MCP server on a local stdin/connection, shielding
the MCP logic from the agent. Their server functionality is merely the API
they expose to local agents.

I think there are two broad models we can discuss: centralized
(framework-based) and distributed (standards-based). Both can coexist
within the ASF. I can very easily **everything** we dp routed by ASF
`llmao` and "tight controlled" by the infrastructure / tooling - but also I
can very easily imagine that ASF contributors and maintainers use their own
LLM /  Subscriptions and local servers, connecting to ASF APIs - this is
IMHO more scalable and more "Apache Way" approach. Both models have pros
and cons, but with the standards published in recent months, the barrier to
using the distributed approach has dramatically lowered.

I personally strongly favor a distributed approach over a centralized one
for MCP servers, SKILLs, and Agent definitions. This allows each component
to be maintained and published separately by different PMCs directly in
their GitHub repos. It is sufficient for their code to be published
somewhere for the MCP server to be installed and run locally. For example,
the MCP servers for Ponymail and Apache Project MCPs in Comdev —which we
use in Magpie—are both simply installed and updated from the Comdev GitHub
repo.

Standards for all of this were established in recent months (MCP 2.0 and
Agents Marketplace standards established in July/August).

We recently converted Magpie to be installed from a distributed marketplace
in preparation for Magpie's 0.2.0 (or maybe even 1.0.0) release. ) release.
I am very happy with the test results; the open standards work smoothly
across many agent harnesses and are already quite powerful.

For example, I used (rather than copying as I was doing so far) the threat
model skills from Alpha-Omega to handle threat modeling for our scan
preparation. I plan to make this accessible for all PMCs and non-ASF
projects—more details will be shared in Glasgow, where we plan to run
workshops addressing PMC needs around managing security issue volume
(including those from our scans).

Moving forward, I plan to pursue what we previously discussed on the Magpie
devlist with Comdev and Incubator PMCs: distributing skill families back to
Comdev and Incubator so they can be managed and exposed directly by those
PMCs via distributed marketplaces, rather than hosting them in the Magpie
repository. There are few benefits to centralizing them in one place; it's
much better to have them managed by the respective PMCs and simply "link"
to them.

Similarly, Legal could distribute their licensing, MCP, and SKILL
marketplace - and we could link to it from Magpie and expose the
"licencing" family of skills.

This approach - I think - supports better scaling and distributes
governance in the true "Apache Way," giving each PMC autonomy over
implementation languages, scope, and publishing methods using open
standards.

Best,
Jarek

On Tue, Sep 8, 2026 at 11:45 AM Jean-Baptiste Onofré <jb...@nanthrax.net>
wrote:

> That's a valid point: magpie is definitely worth a try on podlings.
>
> Regards
> JB
>
> On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com>
> wrote:
> >
> > HI,
> >
> > On the AI side, Apache Magpie has a release skill that will find issues
> in release candidates for you. I suggest that prodlings try this out. Also,
> the Policy MCP can be used to check releases. Having teh Policy MCP make it
> consume a lot fewer tokens, faster, and up to date if the policy does
> change.
> >
> > Kind Regards,
> > Justin
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> > For additional commands, e-mail: general-help@incubator.apache.org
> >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>
>

Re: [DISCUSS] Better License Verification Method

Posted by Jean-Baptiste Onofré <jb...@nanthrax.net>.
That's a valid point: magpie is definitely worth a try on podlings.

Regards
JB

On Tue, Sep 8, 2026 at 8:12 AM Justin Mclean <ju...@classsoftware.com> wrote:
>
> HI,
>
> On the AI side, Apache Magpie has a release skill that will find issues in release candidates for you. I suggest that prodlings try this out. Also, the Policy MCP can be used to check releases. Having teh Policy MCP make it consume a lot fewer tokens, faster, and up to date if the policy does change.
>
> Kind Regards,
> Justin
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
>

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Justin Mclean <ju...@classsoftware.com>.
HI,

On the AI side, Apache Magpie has a release skill that will find issues in release candidates for you. I suggest that prodlings try this out. Also, the Policy MCP can be used to check releases. Having teh Policy MCP make it consume a lot fewer tokens, faster, and up to date if the policy does change.

Kind Regards,
Justin
---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Jean-Baptiste Onofré <jb...@nanthrax.net>.
Hi,

That was exactly my point regarding the PPMC/IPMC vote. Technically,
the PPMC vote is irrelevant; only the IPMC vote is required for a
release to pass. We can view the PPMC vote as a podling validation
prior to the official vote, so a "concurrent vote" is not an issue in
this context (we can totally imagine a podling goes directly to IPMC
vote whatever the PPMC vote is or even occurred).

I am fine with using "improved" wording as long as it does not cause
additional confusion. I would place the emphasis on the IPMC vote
there.

Regards,
JB

On Tue, Sep 8, 2026 at 2:33 AM Greg Stein <gs...@gmail.com> wrote:
>
> On Mon, Sep 7, 2026 at 7:05 PM Dave Fisher <wa...@apache.org> wrote:
> >...
>
> > Incubator podlings simply have no policy route to an “Expedited Release”.
> > None, allowing that would be a new discussion. Expedited Releases must
> > occur on private@ and require informing the board afterwards.
> >
> > If we need a name then how is “Compressed Release”?
> >
>
> Maybe call it an "Incubator-Approved Release"? We're talking about the IPMC
> (the formal PMC constructed per the Bylaws) approving a release. What more
> needs to be done?
>
> Historically, anybody can make a release once they get (3) +1 votes on that
> release. I see no reason to treat the IPMC any differently from the other
> 200 PMCs at the Foundation, just because the repository being released has
> been labeled as a "podling". The Foundation does not recognize that
> construct in its Bylaws, nor any Resolutions I'm aware of, which formalize
> the podling concept as an independent entity.
>
> I believe Tison is correct here, in saying that (3) IPMC Members may vote
> for a podling release to be made, before waiting for a PPMC vote.
>
> Not sure on the name. As all releases are "Incubator-Approved".
> "Compressed" seems fine, but doesn't quite land properly. It also sounds
> like the "faster than 72h" rule or somesuch. "Overlapped Approval" or
> something? Blue-skying.
>
> So needless to say, I asked Grok. It provided a great naming and policy
> guidance:
>
> --
>
> Keep the release name ordinary (Apache Foo 1.2.3). Name the *procedure*:
>
> A release vote may close as *passed by concurrence* when the first-stage
> thread has both (a) a passing first-body tally and (b) at least three
> binding +1 votes from the second body. In that case the second-stage vote
> is not opened.
>
> RESULT subject:
>
> [RESULT][VOTE] Release Apache Foo 1.2.3-rc2 — passed by concurrence
>
> Body line:
>
> Binding second-body votes on this thread: John, Jane, Tom. Second-stage
> vote waived.
>
> --
>
> >...
>
> Cheers,
> -g

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by Greg Stein <gs...@gmail.com>.
On Mon, Sep 7, 2026 at 7:05 PM Dave Fisher <wa...@apache.org> wrote:
>...

> Incubator podlings simply have no policy route to an “Expedited Release”.
> None, allowing that would be a new discussion. Expedited Releases must
> occur on private@ and require informing the board afterwards.
>
> If we need a name then how is “Compressed Release”?
>

Maybe call it an "Incubator-Approved Release"? We're talking about the IPMC
(the formal PMC constructed per the Bylaws) approving a release. What more
needs to be done?

Historically, anybody can make a release once they get (3) +1 votes on that
release. I see no reason to treat the IPMC any differently from the other
200 PMCs at the Foundation, just because the repository being released has
been labeled as a "podling". The Foundation does not recognize that
construct in its Bylaws, nor any Resolutions I'm aware of, which formalize
the podling concept as an independent entity.

I believe Tison is correct here, in saying that (3) IPMC Members may vote
for a podling release to be made, before waiting for a PPMC vote.

Not sure on the name. As all releases are "Incubator-Approved".
"Compressed" seems fine, but doesn't quite land properly. It also sounds
like the "faster than 72h" rule or somesuch. "Overlapped Approval" or
something? Blue-skying.

So needless to say, I asked Grok. It provided a great naming and policy
guidance:

--

Keep the release name ordinary (Apache Foo 1.2.3). Name the *procedure*:

A release vote may close as *passed by concurrence* when the first-stage
thread has both (a) a passing first-body tally and (b) at least three
binding +1 votes from the second body. In that case the second-stage vote
is not opened.

RESULT subject:

[RESULT][VOTE] Release Apache Foo 1.2.3-rc2 — passed by concurrence

Body line:

Binding second-body votes on this thread: John, Jane, Tom. Second-stage
vote waived.

--

>...

Cheers,
-g

Re: [DISCUSS] Better License Verification Method

Posted by Dave Fisher <wa...@apache.org>.
> On Sep 7, 2026, at 4:37 PM, tison <wa...@gmail.com> wrote:
> 
>> IPMC vote (release) SHOULD still follow the release policy defined in
> 
> As this uses the uppercase form, it is a MUST.
> 
> Best,
> tison.
> 
> tison <wa...@gmail.com> 于2026年9月8日周二 07:30写道:
>> 
>> Hi PJ,
>> 
>> I recorded your vote.
>> 
>> To keep the vote thread focused and keep this specific idea discussion
>> focused, I am starting a new thread.
>> 
>> Let's see how we can improve the license/release verification method.
>> 
>> Comment inline.
>> 
>>> -1 binding
>>> 
>>> I would prefer that only podlings that have done a clean release in
>>> the past can take advantage of expedited releases. I'm not against
>>> some mechanism to support shorter release turnarounds but I can't
>>> support the current proposal.
>> 
>> Technically speaking, the concurrent PPMC and IPMC vote is not an
>> expedited release as defined in [1]. It is still a formal and normal
>> vote by either the PPMC or the IPMC.

Incubator podlings simply have no policy route to an “Expedited Release”. None, allowing that would be a new discussion. Expedited Releases must occur on private@ and require informing the board afterwards.

If we need a name then how is “Compressed Release”?

>> 
>> [1] https://www.apache.org/legal/release-policy.html#expedited-releases
>> 
>>> Or the release checklist should be expanded to include items like
>>> whether the source licenses have been validated - current wording
>>> typically just asks that the LICENSE and NOTICE are in the release.
>> 
>> I totally agree that we can extend the release criteria to help
>> podlings make better releases. This item, IMO, can also apply to TLPs.
>> 
>> Actually, the Incubator defines release guidelines in [2], and the
>> IPMC vote (release) SHOULD still follow the release policy defined in
>> [3], which states:
>> 
>>> Every ASF release MUST comply with ASF licensing policy. This requirement is of utmost importance and an audit SHOULD be performed before any full release is created. In particular, every artifact distributed MUST contain only appropriately licensed code per Apache Licensing Policy [4].

I would like to discuss how we can make release candidates and the critical checks more like a periodic step done during development to track progress towards the next release. I’ve been thinking about this with respect to ATR and would like to chat about this during CoC in Glasgow.


>> 
>> [2] https://incubator.apache.org/policy/incubation.html#releases
>> [3] https://www.apache.org/legal/release-policy.html#licensing
>> [4] https://www.apache.org/legal/resolved
>> 
>> So I think it is actually included. The wording could be improved, and
>> proposals are welcome.
>> 
>> But I can try to understand your concerns and consider whether we have
>> better methods to validate source licenses. So far, RAT can do
>> something but it would miss others that are not mechanically wrong.
>> ScanOSS can do some other things but it can still hardly replace human
>> checks.
>> 
>> What are human checks? They are defined as PPMC/IPMC votes cast on the
>> voting thread.

The checks on releases on ATR include more than the standard ones. We do check for binaries in source artifacts for example and what rat excludes you have or concerns you get are available. Perhaps we are closer to giving humans the best information about where to look. Tooling has a workshop on ATR scheduled for the conference.

>> 
>> So technically, even if we repeat the requirement in the guidelines,
>> it doesn't change the practices.
>> 
>> The significant part is to enable and help PPMC/IPMC members verify
>> source licenses more effectively. And I find that the current AI agent
>> can reduce the workload (but you should not outsource the work).
>> 
>> My prompt is in Chinese, but it's as simple as asking the agent to
>> verify all the sources by following our policy, and it would search
>> the Internet to find similar sources. You can even ask for a lower
>> false-negative output in trade-off for a higher false-positive rate,
>> since false positives can be resolved by RM's answers without
>> canceling the vote. As examples, some recent -1 votes [5][6] are
>> spotted with the help of an agent.
>> 
>> [5] https://lists.apache.org/thread/bwr06m0mczd69tnq7h9opm1hnwhn4o9h
>> [6] https://lists.apache.org/thread/8tg6lwh154f5fxvjjkrpdlnxd9l9n37y
>> 
>> If you have other good practices or methods, feel free to share.
>> 
>> Our goal should be to help podlings release better, and waiting a few
>> more hours doesn't magically achieve that. Finally, we need a
>> combination of better methods and better education so that (P)PMC can
>> follow the policy more proactively and more easily.
>> 
>>> Convenience Binaries and their licenses/notices are not always considered.
>> 
>> I'm not sure what you want to emphasize here. Could you elaborate a bit?

Best,
Dave


>> 
>> Best,
>> tison.
> 
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
> For additional commands, e-mail: general-help@incubator.apache.org
> 


---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org


Re: [DISCUSS] Better License Verification Method

Posted by tison <wa...@gmail.com>.
> IPMC vote (release) SHOULD still follow the release policy defined in

As this uses the uppercase form, it is a MUST.

Best,
tison.

tison <wa...@gmail.com> 于2026年9月8日周二 07:30写道:
>
> Hi PJ,
>
> I recorded your vote.
>
> To keep the vote thread focused and keep this specific idea discussion
> focused, I am starting a new thread.
>
> Let's see how we can improve the license/release verification method.
>
> Comment inline.
>
> > -1 binding
> >
> > I would prefer that only podlings that have done a clean release in
> > the past can take advantage of expedited releases. I'm not against
> > some mechanism to support shorter release turnarounds but I can't
> > support the current proposal.
>
> Technically speaking, the concurrent PPMC and IPMC vote is not an
> expedited release as defined in [1]. It is still a formal and normal
> vote by either the PPMC or the IPMC.
>
> [1] https://www.apache.org/legal/release-policy.html#expedited-releases
>
> > Or the release checklist should be expanded to include items like
> > whether the source licenses have been validated - current wording
> > typically just asks that the LICENSE and NOTICE are in the release.
>
> I totally agree that we can extend the release criteria to help
> podlings make better releases. This item, IMO, can also apply to TLPs.
>
> Actually, the Incubator defines release guidelines in [2], and the
> IPMC vote (release) SHOULD still follow the release policy defined in
> [3], which states:
>
> > Every ASF release MUST comply with ASF licensing policy. This requirement is of utmost importance and an audit SHOULD be performed before any full release is created. In particular, every artifact distributed MUST contain only appropriately licensed code per Apache Licensing Policy [4].
>
> [2] https://incubator.apache.org/policy/incubation.html#releases
> [3] https://www.apache.org/legal/release-policy.html#licensing
> [4] https://www.apache.org/legal/resolved
>
> So I think it is actually included. The wording could be improved, and
> proposals are welcome.
>
> But I can try to understand your concerns and consider whether we have
> better methods to validate source licenses. So far, RAT can do
> something but it would miss others that are not mechanically wrong.
> ScanOSS can do some other things but it can still hardly replace human
> checks.
>
> What are human checks? They are defined as PPMC/IPMC votes cast on the
> voting thread.
>
> So technically, even if we repeat the requirement in the guidelines,
> it doesn't change the practices.
>
> The significant part is to enable and help PPMC/IPMC members verify
> source licenses more effectively. And I find that the current AI agent
> can reduce the workload (but you should not outsource the work).
>
> My prompt is in Chinese, but it's as simple as asking the agent to
> verify all the sources by following our policy, and it would search
> the Internet to find similar sources. You can even ask for a lower
> false-negative output in trade-off for a higher false-positive rate,
> since false positives can be resolved by RM's answers without
> canceling the vote. As examples, some recent -1 votes [5][6] are
> spotted with the help of an agent.
>
> [5] https://lists.apache.org/thread/bwr06m0mczd69tnq7h9opm1hnwhn4o9h
> [6] https://lists.apache.org/thread/8tg6lwh154f5fxvjjkrpdlnxd9l9n37y
>
> If you have other good practices or methods, feel free to share.
>
> Our goal should be to help podlings release better, and waiting a few
> more hours doesn't magically achieve that. Finally, we need a
> combination of better methods and better education so that (P)PMC can
> follow the policy more proactively and more easily.
>
> > Convenience Binaries and their licenses/notices are not always considered.
>
> I'm not sure what you want to emphasize here. Could you elaborate a bit?
>
> Best,
> tison.

---------------------------------------------------------------------
To unsubscribe, e-mail: general-unsubscribe@incubator.apache.org
For additional commands, e-mail: general-help@incubator.apache.org