On 10/30/17 17:30, Pramod Immaneni wrote:
On Sat, Oct 28, 2017 at 7:47 AM, Vlad Rozov <vro...@apache.org> wrote:
Don't we use unit test to make sure that PR does not break an existing
functionality? For that we use CI environment that we do not control and do
not introduce any changes to, but for example Apache infrastructure team
may decide to upgrade Jenkins and that may impact Apex builds. The same
applies to CVE. It is there to prevent dependencies with severe
vulnerabilities.
Infrastructure changes are quite different, IMO, from this proposal. While
they are not in our control, in majority of the cases, the changes maintain
compatibility so everything on top will continue to run the same. In this
case a CVE will always fail all PRs, the code changes have nothing to do
with introducing the CVE. I did make the exception that if a PR is bringing
in the CVE yes do fail it.
There were just two recent changes, one on Travis CI side and another on
Jenkins side that caused builds for all PRs to fail and none of them was
caused by code changes in any of open PRs, so I don't see how it is
different.
A code change may or may not have relation to CVE introduced. For newly
introduced dependencies, there may be known CVEs. In any case I don't
think it is important to differentiate how CVE is introduced, it is
important to eliminate dependencies with known CVEs.
There is no "stick" in a failed build or keeping PR open until dependency
issue is resolved or unit test failure is fixed. Unless an employer
punishes its employee for not delivering PR based on that vendor priority,
there is no "stick". As we already discussed, the community does not have a
deadline for a PR merge or for a release to go out. In a case there is a
problematic dependency (with CVE or category X license) you as a PMC
suppose to -1 a release (at least I will). Will you consider -1 as a
"stick"? For me, it is not about punishing an individual contributor, it is
a priority and focus shift for the entire community, not a "stick" for an
individual contributor.
The stick I am referring to is failing all PRs hoping that will make people
address CVEs. It's got nothing to do with an employer, people contributing
to open source can't expect they can control what the outcome will be or
what form it will take. You may be confusing this with some other issue. In
some of the arguments, you are assuming this scenario is similar to build
failures from failing unit tests and I am arguing that premise. I don't
think we should bring regular development to a halt whenever a matching CVE
is discovered, unless there is some other secondary reason like merging a
PR will make it difficult for a CVE fix to be made. Have you given a
thought to what I said about having a separate build that will notify about
CVEs.
As I mentioned, there is no stick, no deadlines and no issues keeping
PRs open until builds can be verified on CI environment. Fixing a failed
build is a priority for the *community* not a stick for an individual
contributor.
I don't see why keeping PRs open (for whatever reason) brings regular
development to a halt. Nobody is going to put github repo offline.
Contributors may continue to open new PR, collaborate on existing PRs
and add more changes (and need to be patient for those changes to be
reviewed and accepted). The regular development will continue with the
only exception that the next commit to be merged must address the build
issue (whether it is a failed unit test or newly found CVE).
I don't see much value in a separate build and do not plan to put effort
in that direction. Additionally, will not a separate build that only
checks for CVE will trigger your initial concern of disclosing CVE in
public?
Thank you,
Vlad
On 10/27/17 14:28, Pramod Immaneni wrote:
On Fri, Oct 27, 2017 at 11:51 AM, Vlad Rozov <vro...@apache.org> wrote:
On 10/26/17 11:46, Pramod Immaneni wrote:
On Thu, Oct 26, 2017 at 10:39 AM, Vlad Rozov <vro...@apache.org> wrote:
I guess you are mostly concerned regarding new CVE in an existing
dependency. Once such CVE is added to a database, will it be better to
know
about it or postpone discovery till we cut release candidate? In case
it
is
reported only during release cycle, there is much less time for the
community to take an action and it still needs to be taken (as a PMC
member
you are responsible for preventing release with severe security issue
going
out). If it is reported once the information becomes available, there
is
more time to research and either upgrade the dependency with newly
found
CVE, agree that it does not impact the project.
This would be the more commonly occurring scenario. We can always know
about the CVEs because of your changes. We don't need to fail builds to
do
that. I am not asking you to remove the reporting. There is no set time
for
a release so having less time during release for addressing relevant
CVEs
does not come up. There is also nothing preventing anyone from looking
at
these reports and taking action earlier.
I don't see why it will be more commonly occurring scenario, but I think
it is equally important to prevent new dependency with severe
vulnerabilities being introduced into the project and check existing
dependencies for newly discovered severe vulnerabilities.
How will we know about CVE if it is removed from CI build? Why require
manual verification when it can be done during CI build and does not
affect
builds done by individual contributors?
While there is no set time for a release, there is no set time for a PR
merge as well.
Yes, nothing prevents anyone from looking at the dependency
vulnerabilities, but there is a better chance that "anyone" does not mean
nobody if CI build fails.
I still do not understand why you value an individual contributor and PR
over the community and the project itself. Once there is a severe
security
vulnerability, it affects everyone who cares about the project,
including
all contributors. I don't see a problem with a PR being in a pending
(not
merged) open state till a build issue is resolved.
That is a mischaracterization that you have stated before as well. A
project cannot grow without contributions and without policies that
create
a supportive enviroment where that can happen, I don't see the need to
put
unnecessary obstacles for legitimate contributions that are not the
cause
of a problem. Everytime there is a matching CVE the PRs are going to get
blocked till that CVE is addressed and I am not confident we have the
bandwidth in the community to address this expediently. It is also
inaccurate to equate this to PR not being merged till build issues are
resolved as it derives from an assumption that CVE is same as a build
failure.
While project can't grow without individual contributions, project is a
result of a large number of contributions from a number of contributors.
Some of those contributors are not active anymore and will not provide
any
fixes should a vulnerability be found in their contribution. It becomes a
shared responsibility of all currently active community members and those
who submit PR are part of the community and share that responsibility,
are
not they? If a contributor considers him/herself as part of a community,
why he or she can't wait for the build issue to be resolved or better
take
an action on resolving the issue? The only possible explanation that I
see
is the one that I already mentioned on this thread.
If you see this as unnecessary obstacles for legitimate contributions,
why
to enforce code style, it is also unnecessary obstacle. Unit test? Should
it be considered to be optional for a PR to pass unit tests as well? What
if an environment change on CI side causes build to fail similar to what
happened recently? Should we disable CI builds too and rely on a
committer
or a release manager to run unit tests? If CI build fails for whatever
reason, how can you be sure that if it fails for another PR as well, that
they both fail for the same reason and there is no any other reasons that
will cause a problem with a PR?
I don't know how failing PRs because of CVE, which we don't introduce,
don't control, no idea of and possibly unrelated would fall in the same
bucket as unit tests. I am at a loss of words for that. There is no reason
to block legitimate development for this. This can be handled separtely
and
in parallel. Maybe there is a way we can setup an independent job on a
build server that runs nightly, fails if there are new CVEs discovered and
sends an email out to the security or dev group. You could even reduce the
CVE threshold for this. I don't believe in a stick approach (carrot and
stick metaphor) and believe in proportional measures.
Thank you,
Vlad
On 10/26/17 09:42, Pramod Immaneni wrote:
On Thu, Oct 26, 2017 at 12:08 AM, Vlad Rozov <vro...@apache.org>
wrote:
There is a way to add an exception, but it needs to be discussed on a
case
by case basis. Note that CVEs are not published until a fix is
available.
For severity 8 CVEs I expect patches to become available for the
reported
version unless it is an obsolete version in which case, the upgrade
to
a
supported version is already overdue.
I think we should retain the ability to make that choice of what and
when
to upgrade rather than hard enforce it. Like I mentioned the CVE may
not
apply to us (it has happened before), even though it may be beneficial
upgrade generally when its not applicable, there may not be the
bandwidth
in community to do the necessary changes to upgrade to a newer version
especially if those dependencies don't follow semver (has happened
before
as well, remember effort with ning). My caution comes from
experiencing
this situation before.
I don't see how reporting helps. If a build succeeds, I don't expect
anyone to look into the report, it is only when CI build fails,
committers
and reviewers look into the details.
We can add a mandatory step during release that we need to assess
CVEs
matching this criteria before proceeding with the release. This could
end
up requiring upgrade of some dependencies and in other cases it may
not
be
needed. This assessment can also happen more often in adhoc fashion
offline
before release based upon interest from community. I am also open to
making
it mandatory with every PR, in future, like you are suggesting, if we
see
sufficient uptake in community on these issues. From experience this
is
not
there currently and hence I don't want to do this now.
IMO, it does not matter how CVE is introduced. It may be a new
dependency
with an existing CVE or it can be a new CVE for an existing
dependency.
In
both cases, dependency with the CVE needs to be fixed.
In one case the PR is not directly responsible for the issue and
hence
we
should avoid penalizing them or block them.
Thank you,
Vlad
On 10/25/17 11:58, Pramod Immaneni wrote:
Thanks that sounds mostly fine except what happens if there is a cve
matching that severity in a dependency but it doesnt affect us
because
let's say we don't exercise that part of functionality *and* there
isn't a
fix available or there is a fix but the upgrade requires significant
effort
(for example if we need to move to a new major version of the
dependency
or
something like that). Is there a way to add an exception like we did
for
checkstyle in the interim. How about reporting instead of failing
the
builds. One of the steps in release process could be to investigate
these
reports before proceeding with the release. If a PR introduces new
cves
by
virtue of adding a new dependency or changing an existing version,
that
would be of interest and can be subject to failure. Is there a way
to
distinguish that?
On Wed, Oct 25, 2017 at 8:52 AM Vlad Rozov <vro...@apache.org>
wrote:
A CVE (should there be a vulnerability in existing or a newly
introduced
dependency) will not be exposed during the CI build, but the build
will
fail if the CVE has severity 8 or above. To get the details, it
will
be
necessary to run dependency check manually.
Thank you,
Vlad
On 10/24/17 16:27, Pramod Immaneni wrote:
There was a lot of discussion on this but looks like there was no
final
agreement. Can you summarize what your PR does? Are we disclosing
the
actual vulnerabilities as part of the automated build for every
PR?
That
would be a no-no for me. If it is something that requires manual
steps,
for
example as part of a release build, that would be fine.
On Mon, Oct 23, 2017 at 1:16 PM, Vlad Rozov <vro...@apache.org>
wrote:
Please see https://github.com/apache/apex-core/pull/585 and
APEXCORE-790.
Thank you,
Vlad
On 9/14/17 09:35, Vlad Rozov wrote:
Do you expect anything else from the community to recognize a
contribution other than committing it to the code line? Once
there
is
a
steady flow of quality contributions, the community/PMC will
recognize
a
contributor by making that contributor a committer.
Thank you,
Vlad
On 9/12/17 13:05, Sanjay Pujare wrote:
For a vendor too, quality ought to be as important as security
so
I
don't
think we disagree on the cost benefit analysis. But I get your
drift.
By "creative incentive" I didn't imply any material incentive
(although a
gift card would be nice :-)) but more along the lines of what a
community
can do to recognize such contribution.
Sanjay
On Tue, Sep 12, 2017 at 8:10 AM, Vlad Rozov <vro...@apache.org>
wrote:
I guess we have a different view on the benefit and cost
definition.
For
me the benefit of fixing CI build, flaky unit test, severe
security
issue
is huge for the community and is possibly small (except for a
security
issues) for a vendor.
By "creative" I hope you don't mean that other community
members,
users
and customers send a contributor a gift cards to compensate for
the
cost
:). For me PR that is blocked on a failed CI build is
sufficiently
incentive for a contributor to look into why it fails and fixing
it.
Thank you,
Vlad
On 9/11/17 23:58, Sanjay Pujare wrote:
I don't want to speak for others and I don't want to
generalize.
But
an
obvious answer could be "cost-benefit analysis".
In any case we should come up with a creative way to
"incentivize"
members
to do these tasks.
