SF Open Source Voting TAC

Draft Meeting Minutes: November 8, 2018

Order of Business

1. Call to Order & Roll Call

Chair Jerdonek called the meeting to order at 6:03 p.m. Present: Members Jerdonek, Kattouw, and Wasserman. Members Hage and Philips arrived late (excused) during item #3 at 6:17 p.m. and 6:18 p.m., respectively. Also present: Secretary Chan.

2. General Public Comment

None.

3. Open Source Voting Project Plan

Chair Jerdonek welcomed City CIO and Executive Director of the San Francisco Department of Technology, Linda Gerull. Committee members had questions regarding the draft plan she presented (included in the agenda packet).

4. Approval of Minutes of Previous Meeting

Member Kattouw noted a typo and with that correction, moved to approve the minutes of the September 13, 2018 meeting. Member Hage seconded. Upon voice vote, the motion carried unanimously 5-0.

5. Administration

Chair Jerdonek introduced new TAC member Brandon Philips, who gave a short self-introduction.

The decision was to have the next committee meeting on December 6 rather than December 13.

Member Kattouw said he can speak on behalf of the committee at the next Commission meeting on November 21.

Public comment:

Mr. David Cary complimented the committee and asked if they had addressed the Civil Grand Jury’s proposal to create a portal or site for the open source project. Chair Jerdonek referred him to the project page link on the committee’s website whose purpose was to fulfill that recommendation.

6. Member Reports

Member Kattouw reviewed some of the comments made at the Commission meeting regarding the Department of Technology’s (DT) draft Project Initiation Plan. He was pleased by the responsiveness to the comments he raised at the last OSVTAC meeting. He also spoke about his experiences observing the Logic and Accuracy testing held by the Department. He said he observed the elections results reporting and was generally pleased with how it was done. He heard that Vancouver has a process similar to New Hampshire where QR codes are created before the ballot is scanned. He said he is looking forward to when we will have CVR data.

Member Hage reported on his elections work, which includes writing programs to convert election results for thirty-five counties. He spoke about the fact that the State collects all county voting information but only uses some of it and throws the rest away. None of it is available to the public. There was a short discussion about Member Hage‘s request for documentation or manuals for San Francisco’s current Dominion voting system and whether this could distract from or harm the committee’s work.

Member Wasserman reported on his keynote talk on open source in Shenzhen, China.

Chair Jerdonek reported that he sent to Director Gerull the questions that TAC developed and voted on at the previous meeting. He was a polling place inspector on Election Day. At his polling place there was a larger ballot discrepancy during the reconciliation than he had seen in the past. This was due in part to unused vote-by-mail ballots being counted as voided precinct ballots. He said he will tell the committee when the Department does the random selection of precincts if they want to observe.

He also mentioned a panel discussion he participated in, along with Ben Adida, the creator of the online voting system Helios.

Public comment:

Mr. David Cary commented that it is specifically in the area of verifiable vote counting that open source exceeds proprietary closed source systems’ credibility. It is more important to get a core system up and running and add extra capabilities only later.

7. Contributor License Agreements (CLA’s) for OSVTAC Projects

Member Kattouw researched this topic, taking into account previous committee discussions on issues related to licensing, intellectual property, etc. He summarized that the Apache CLA is simple and allows the entity you are agreeing with to re-license your contribution. For GPLv3 projects this is a concern. He didn’t like the JS Foundation CLA, which does not include patent grants. He favored one used by the Harmony Project (harmonyagreements.org). All of this was covered in his email (see agenda packet).

The issue was also raised regarding the City being the owner of the project and protecting itself from possible complications from outside contributors to it or future revisions. Any such action might require action on the part of the Board of Supervisors. If ownership does not reside with the City, then the CLA might not be necessary. The City Attorney’s office would need to confirm, but in initial discussions the Deputy City Attorney felt the City is a copyright owner.

Member Philips felt that GPLv3 is the most aggressive copyleft license, whereas more and more organizations are leaning towards licenses like Apache 2.0. Member Philips also mentioned using a DCO (Developer Certificate of Origin) as an alternative to a CLA, which shows up as a “Signed-off-by” line in a commit.

Member Wasserman mentioned that the evolution of GPL came out of a problem Richard Stallman had with “tivoization.” Member Kattouw will revise his proposed CLA with a re-licensing clause, and Chair Jerdonek will review it with the Deputy City Attorney.

Public comment:

Mr. David Cary said his understanding was that for federal regulatory certification, they would have to show origin of software. He wasn’t sure how that applied to the state certification process. He also felt that the Affero GPL is preferable to the GPL.

8. Voting System Component Development

Member Hage reviewed what he has done since the last meeting on the Open Results Reporter (ORR), which uses Python and Jinja templates. He asked Member Kattouw to write some CSS templates. Member Kattouw said he would have this done by the January meeting. Member Hage said the code in ORR could also be used to generate accessible sample ballots. This could replace the service the City uses now (OmniBallot Online).

Chair Jerdonek reviewed some of ORR’s online documentation and the sample demo output, which is available online. It already has certain reporting features that the City currently doesn’t have. He thought the next step could be to have the data from the November election, a script to generate from that data the input files for ORR, and an attractive template to present ORR’s output.

He wanted to explore how the committee could expand and better organize its development of the Open Results Reporter – given “sunshine” restrictions – and see if the committee could work on more than just the results reporting component.

Member Wasserman wanted to explore how the committee could actually get this module integrated into the final project rather than just having it as an academic exercise. Member Wasserman felt that it would be good to take the “prettified” ORR, give it to DT, and see how they respond.

The discussion of how to proceed considered individual members working on their own but left open how to keep everyone on the same track. There was also the question of how public contributions could fit in. Member Hage thought it would help to explain how working in a container environment and using digital signatures could help address security.

Chair Jerdonek explained to Member Philips how the ORR accepts tabulated vote totals in JSON and TSV forms, and uses template files to format them for public viewing.

Chair Jerdonek said he will continue to flesh out features of the results reporter. He asked Member Hage if he could create a script to generate from the real election data the input files for ORR, and to store those input files in the sample data repo. He also asked if there could be a regression test to check that the conversion code is generating the input files correctly, as the conversion code changes.

Public comment:

Mr. David Cary suggested the route of taking some of the concerns DT may have about the current voting system and constructing the TAC project in a way that addresses those questions. This could also provide proof of how it can fit into the larger project goals without having to wait for planning out the entire system at the beginning.

9. Committee Recommendations

Member Hage suggested going back to the section on security and expanding the background appendix, before DT gets too far along.

Member Wasserman repeated that he thought the committee’s recommendations should be kept in the mind of DT to ensure the parameters are maintained. He asked if discussion should be focused on using containers or in a broader sense.

Member Hage said he will work more on the glossary to amplify terms for cryptography and security. Member Wasserman suggested a way to present this via identifying the vulnerabilities and then offering the solution to each. Chair Jerdonek said that definitions didn’t need to be expansive or very technical.

Member Kattouw thought if Member Hage separated the topic of security into two parts: (1) to verify the security of the data flow (what is inputted is outputted correctly), and (2) to ensure that the software you think is running the machine IS running it, it would read more easily.

10. Topics for future discussion

The topics were the topics that were mentioned earlier in the meeting: security, and continued discussion of CLA’s.

Adjourned at 9:10 p.m.