Present: Maarten L, Tom D (Notes), David K, Matt Doidge, Federica A, Roberta M, Patrick, John SDSJr, Dave D, Berk B
Apologies: Mischa S, Enrico V
Notes:
Action: Maarten to look at reviving the RTE Task Force
- Usage of SHA1, which should be phased out for a while now
- https://twiki.cern.ch/twiki/bin/view/LCG/EL9vsSHA1CAs
- SHA1 not supported on EL10 - cannot just enable SHA1, but you must enable legacy settings, which comes with several other things
- It is not unanticipated that this would be blocked by policy levels
- Therefore, should look to move away from SHA1 where possible
- Maarten & Brian have discussed
- Most middleware shouldn't worry about SHA1 in root certs, as they exist in the Trust store.
- Currently verifying whether this works - looking at dCache (would also cover StoRM, as both use CaNL-Java). Brian has also done other checks, so HTCondor has likely also been checked, but XRootD by default will also check the root certificates - Maarten has asked Brian for details about what sites would need to check or configure.
- Dave K contributes from IGTF (EuGridPMA & TAGPMA)
- IGTF is sympathetic, but has not set a deadline
- Pointed out that if the CA root certificate is the only having SHA1, then it doesn't matter because it's in the store
- DigiCert to be suspended from TAGPMA due to not engaging
- Australian Site in Melbourne having issues with their host and user certificates, formerly issued by DigiCert
- after the meeting, Maarten has updated the relevant GGUS ticket with a pointer to the ASGC catch-all CA for Asia-Pacific
- No Catch-all CA for US & EU
- Dave K expresses that EU doesn't need one due to other routes
- eMudhra CA would also be an option - commercial CA in India
- Look to work toward no longer depending on our own set of CAs, which these days are still needed because of our user (not host) certificates
- Need to, as a community, review what is required of CAs to ensure trustworthiness
- We might need to be concerned about "whims" of the CA/Browser Forum impacting us badly
- But then again, if those CAs work for the rest of the internet, why should we remain an exception?
- Google Trust Service CAs are nearly now approved by TAGPMA
- Let's Encrypt is also expected to get there next year
- WLCG may need to set deadlines in order to see movement with this
- Had been suggested already in WLCG Ops
- End of 2026 has been suggested, which would then need to be announced soon
- At the very least, the few SHA-1 CRLs should be changed to SHA-2
- Separate issues around CRLs which for many CAs are not available via IPV6
- there is a workaround, with an IVP6 mirror
- Various "Dark Clouds" around Certs, means token work must keep progressing steadily
Job Accounting with only Tokens
- Complete solutions for ARC-CE and for HTCondor CE with HTCondor batch system
- Not a solution when other batch systems are used instead
- HTCondor solution is orthogonal to APEL's own solution, but decoupled from APEL code
- Look to imitate this within the APEL code, to provide a solution.
- Maarten to look to have a direct meeting with Adrian to look and make a plan of action
- Look so that other batch systems - including Slurm - could hopefully be covered the same way.
AOB:
- Closed Token Task Force progressing with WLCG issues, but still needs AuthZ group to do some of the actual work!
- Also other groups, such as Token Trust & Traceability, DOMA BDT
- AuthZ WG is mainly concerned with the token profiles and IAM evolution
- Tentative timeline with some deadlines over the next year
- Maarten presented on the timeline recently, and notes that further progress following the new profile should be expected over the course of this year.
- Updated tentative timeline is shown on page 18 of this talk
- Mind the 2025 typo in the milestones for 2026
- TTT Group working on a token risk assessment since the middle of summer, and should report on this by the end of the year.
- Questions around status of no longer storing tokens
- Noting concerns around DB handling ever more concurrent tokens (21 M for ATLAS, uncharted territory for IAM).
- Aim is to do this in v1.14 after required DB changes are done in v1.13, so that roll back would be more manageable, if needed
- Federica to follow up with Francesco & Enrico on the possibility of a v1.14 pre-release to be used for ATLAS.
- Token issuer public key cache standardization proposal
- Dave D notes this is waiting on Brian B, and shares that as Brian has a proposal due, this may be delayed.
There are minutes attached to this event.
Show them.