Markdown Version | Session Recording

Session Date/Time: 10 Jan 2023 19:00

TOOLS

Summary

The TOOLS working group discussed several ongoing infrastructure projects, including the planned migration of DNS services to Cloudflare, recent OpenSuSE upgrades, and a significant delay in the PostgreSQL database migration due to case-insensitivity issues. Other key topics included how to handle missing historical metadata for older drafts (Issue 4933) and the sensitive topic of providing public APIs for lists of email addresses from mailing list posters. Updates were also provided on various development projects like the Wiki migration, Datatracker, XML2RFC, and the Note Well display.

Key Discussion Points

• DNS Configuration Migration to Cloudflare:
  • Agreement with Cloudflare under Project Galileo to host DNS.
  • Ability to specify secondary DNS servers to continue using existing volunteer secondaries.
  • An outstanding question regarding how zone transfers to secondaries would point to Cloudflare edge servers vs. origin servers; awaiting clarification from Cloudflare.
  • Targeting a transition this month, possibly early February, but Robert will be offline Jan 30 - Feb 3.
  • The transition plan involves at least one, possibly a couple of days, without DNSSEC signing.
• OpenSuSE Upgrades (ietfa, ietfx, RPCs primary):
  • Point upgrades for ietfa and ietfx from OpenSuSE 15.3 to 15.4 were largely smooth.
  • The RPCs primary server is scheduled for the same upgrade next week.
  • One issue identified: the opendkim package had a policy change regarding Unix permissions on a configuration file, causing it to temporarily reject messages. This led to a delay in email delivery to lists, though messages were archived immediately and no data was lost.
• PostgreSQL Migration:
  • A "stopping bug" was found during the production migration to PostgreSQL.
  • The core issue is a difference in collation: MySQL treated many fields as case-insensitive for comparisons (e.g., email addresses), while PostgreSQL is case-sensitive by default.
  • Initial testing underestimated the scope of this difference. If unaddressed, it could lead to conflicts, such as identical-looking email addresses (differing only by case) creating new records.
  • A full code scrub is required to identify and address all queries and references that might be affected by case sensitivity, particularly primary keys on character type fields.
  • This effort is estimated to take several weeks.
• Doc History for Older Draft Versions (Issue 4933):
  • The SSH community identified an issue where links to older versions of drafts point to drafts lacking doc_history records, meaning metadata like working group state or document shepherd is missing for those versions.
  • This is common for many older Internet-Drafts before they became RFCs.
  • Discussion centered on whether to "make up" historical metadata (with negative ramifications) or to change the view behavior to display older draft text (which is available) but explicitly state that metadata for that version is unknown.
  • Robert indicated changing the view to display text with unknown metadata is likely the preferred route but requires significant refactoring.
  • Suggestions were made to represent "unknown data" within the existing metadata schema, potentially with specific values or objects, rather than inventing historical facts.
• Public API for Email Address Lists:
  • Discussion around whether to expose a public API that would make it easier to extract lists of email addresses of individuals who have posted to any mailing list in a given period (e.g., year, month).
  • The motivation is to support internal reporting (e.g., for ISG, IETF annual reports) and make information more transparent.
  • Concerns were raised about making it too easy for spammers to harvest targeted lists (e.g., "all TLS developers").
  • Participants acknowledged that such lists are already technically crawlable from existing archives and the Datatracker API V1, but the "difficulty of extraction" is a key consideration.
  • Russ noted the utility of having current email addresses for legitimate purposes, such as the RFC editor being able to contact authors about errata, which is currently challenging for older RFCs.
  • Suggestions included requiring an API key (even if free and easily obtainable) to add a layer of friction, making it a POST request rather than a simple GET.
  • A sense of those present indicated that the primary concern is legitimate use cases (like contacting authors) versus the difficulty of spam harvesting.
• Note Well Display:
  • Efforts are underway (Greg and Robert) to review where the "Note Well" is displayed to ensure participants are consistently presented with it in key interactions (e.g., ID submission).
  • Legal advice indicated it is not required for IPR declarations.
  • Work also includes informing new mailing list subscribers about the Note Well.
• Meeting Time Zone Policy:
  • A recurring check was made on whether to continue fixing the meeting time in US time (shifting with daylight saving) or fix it to UTC.
  • Not seeing anyone expressing a strong preference, the existing practice will continue.

Decisions and Action Items

• Decision: Reschedule the production migration of the Datatracker to PostgreSQL to April 11th, after the Yokohama IETF, to allow for a full code scrub addressing case-sensitivity issues. An announcement will be sent out.
• Action Item: Robert to obtain further information from Cloudflare regarding zone transfer behavior for secondary DNS and plan the transition date.
• Action Item: TOOLS team to discuss and propose a solution for handling missing doc_history metadata for older drafts (Issue 4933), considering options for representing unknown data.
• Action Item: TOOLS team to develop a proposal for a public API for mailing list poster email addresses, taking into account the discussions on spam concerns vs. legitimate utility.
• Action Item: Jay to provide an update on the infrastructure strategy RFP next week.
• Decision: XML2RFC release 8.95 is planned, focusing on permitting ASCII within T without the use of U. Support for editorial stream name and boilerplate emission is also targeted for this release, though it might defer if blocking points arise.
• Decision: The practice of keeping the TOOLS meeting time fixed in U.S. time (shifting with U.S. daylight saving) will continue due to no strong preference for changing to UTC.

Next Steps

• Continue the Wiki migration to Wiki.js.
• Deploy upcoming Datatracker release (late today or tomorrow).
• Focus Datatracker development cycles on the PostgreSQL transition, IAB/ISOC artifact integration, and publication/editorial stream document life cycle work.
• Organize outstanding issues with the bib XML service and prioritize work with Ribose.
• Address Authortools issues related to recent Ruby releases and spider interactions.
• Continue work on the Wagtail portion of the website.
• Proceed with the planned XML2RFC 8.95 release and subsequent work on other requested features.
• Continue efforts to improve the display and prominence of the "Note Well" in IETF systems.