Hi everyone, Thanks as always to all of today's participants. Here is the meeting summary:
Date: April 30, 2026 Attendance: - Cole Greer - Valentyn Kahamlyk - Lyndon Bauto - Josh Shinavier Groovy Removal from Server: We had a brief discussion of the state of the current PR (https://github.com/apache/tinkerpop/pull/3384) which has now addressed all outstanding feedback. Valentyn raised the ability for providers to extend server Settings as a useful followup to this work. The simplest form of this could simply be adding a generic map into Settings to collect arbitrary provider-specific configs. Plans for OLAP: We broke into a discussion about modernizing our OLAP implementations. Lyndon and Valentyn provided much of the feedback here. The main points raised are that spark remains an ideal foundation for OLAP, however the way we utilize it should be modernized. The 2 main changes discussed were a migration from RDD to Dataframes, and the ability to stream or partially load necessary data into spark for each traversal, without needing to load the entire graph. Valentyn has an old [DISCUSS] thread regarding step interfaces for efficient data retrieval which I believe is relevant to this discussion (https://lists.apache.org/thread/2y62d7h2fhwb7nlbh7z1tm2t35cgpchc). GQL Support: The topic of GQL support was raised, we briefly discussed the work Stephen has recently picked up following the Match GQL proposal driven by YouTrackDB. Valentyn believes there is value in TinkerPop offering full support of GQL beyond just the match step. It's an interesting item to consider in our long-term plans, although it faces considerable complications with regards to items such as query optimization and indexing. Driver Parity: Lyndon asked about feature parity across all of our drivers. I suggested that better parity is definitely a goal for TinkerPop 4, although the exact extent of this remains a bit vague. The biggest feature difference at the moment is the advanced connection pooling and support for multiple endpoints which is exclusive to Java. I would like to see other drivers receive such features, although I would also like to reduce the complexity of the Java driver as long as it doesn't result in a loss of capability. I expect this to be an evolving conversation as we continue progressing towards TinkerPop 4. Interchange Formats: Josh briefly demoed several different json formats for graph data and graph schema interchange, as well as a protobuf format. We discussed mapping these formats to a slightly different data model to the discussed TinkerPop schema data model. I believe these formats should be a good fit for TinkerPop's GraphSchema implementation. There will be more to share and discuss on interchange formats in a future gathering (possibly in the next regular timeslot, or as an adhoc meeting). Thanks again to everyone who took part, the next gathering is scheduled for May 14 at 16:00 UTC. Please reach out if you're interested in participating, or would like to request a different meeting time. Regards, Cole On 2026/04/17 01:31:52 Cole Greer wrote: > Hi everyone, > > Thanks to everyone who was able to join this morning, here are my meeting > notes: > > Date: April 16. 2026 > > Attendance: > - Cole Greer > - Valentyn Kahamlyk > - Yang Xia > - Ken Hu > - Josh Shinavier > - Kris McGinnes > > Technical discussions today covered 2 main topics, groovy deprecation/removal > in the server, and graph data and schema interchange formats: > > > Groovy Deprecation: > > We had a chat about my recent proposal regarding the deprecation of groovy in > the server: https://lists.apache.org/thread/nrnlp4b8vrwbzhr366swn1fzk9wygsmg, > and it's associated PR. > > Valentyn contrasted the WIP TinkerPop solution to Aerospikes own approach to > keep groovy out of the server, as well as suggested additional test cases to > add to the PR. Concerns were raised regarding the groovy script engine still > being available in the server even if it's excluded from the server config > yaml. I'll be taking these concerns and updating the PR accordingly. > > > Interchange Formats: > > We had a brief discussion regarding a desire for a standardized interchange > format for both graph data and graph schemas. Concerns with our existing > formats such as GraphSON included that they are hard to process without using > our provided serializers (only available in our GLV languages), and that it's > hard to decompose and process files in chunks. > > Piggy backing on a standard format such as JSON gives flexibility to work > from any language. Some concerns were raised with CSV due to the > complications of mapping property keys to the csv columns (graph schema > impact the structure of the CSV). > > There was some discussion of how it would be convenient if we could directly > adopt a standard from GQL, or at least adopt a format which is generally > aligned, and provide standardized mappings (possibly lossy) between our graph > data model and that of GQL's. > > Josh intends to lead a more detailed discussion on this topic in the next > gathering. > > > In addition to these main topics, we also broke out into an interesting side > discussion regarding the impact of recent AI development on our individual > workflows. > > Thanks to everyone who took part, the next gathering is scheduled for April > 30. Please reach out if you're interested in participating. > > Regards, > Cole > > On 2026/04/03 02:26:06 Cole Greer wrote: > > Hi everyone, > > > > Thanks to everyone who joined today's gathering, here is the meeting > > summary: > > > > Date: April 2, 2026 > > > > Attendance: > > > > - Cole Greer > > - Kris McGinnes > > - Valentyn Kahamlyk > > - Josh Shinavier > > - Yang Xia > > > > This week the primary topic of discussion was a proposal for graph schema > > interfaces for TinkerPop, which I had developed based on feedback in > > previous gatherings. This proposal can be found here: > > https://gist.github.com/Cole-Greer/335ea5c44d6be1d677a7894f968fadcd, > > although the feedback from today has now been integrated into an updated > > version which has spawned a dedicated [DISCUSS] thread: > > https://lists.apache.org/thread/2nl6z63ldsbdsqr1mczvr4cqt68jtc7w. > > > > The topic of property graph interchange formats was briefly raised, which I > > expect a deeper discussion around new interchange formats for TinkerPop to > > take place in the next gathering on April 16. > > > > Thanks to everyone who attended and provided inputs to the schema proposal. > > > > Thanks, > > Cole > > > > On 2026/03/11 05:02:28 Cole Greer wrote: > > > Hi everyone, > > > > > > My apologies for the delays in posting notes from last week as I was > > > traveling. For those who were unable to attend, we had an in depth > > > discussion outlining each of our individual goals regarding schema > > > support in TinkerPop. We are intending to continue the discussion in the > > > next gathering. I have rescheduled the next meeting to take place on > > > Thursday March 19, at 16:00 UTC (09:00 PDT, 18:00 SAST, 17:00 CET) to > > > re-establish our normal 2-week cadence. > > > > > > Here are the notes from the last gathering: > > > > > > Date: March 5, 2026 > > > > > > Attendance: > > > - Cole Greer > > > - Valentyn Kahamlyk > > > - Josh Shinavier > > > - Yang Xia > > > - Stephen Mallette > > > - Pieter Martin > > > > > > Details: > > > > > > Josh led the initial discussion with some prepared slides > > > (https://docs.google.com/presentation/d/1IHb8rVfvtoQyvvnYgYGNu5aYf45mwGyxtvfI7ILLOfI) > > > on schema in TinkerPop. > > > > > > This discussion explored defining requirements for schema expressiveness > > > and capability in TinkerPop, starting with a type system, and moving on > > > to validation and constraints. The slides presented many features for a > > > very rich and capable type system, some of which those in attendance > > > generally agreed may be more than some providers are willing to support. > > > It is expected that providers will need to be given the ability to opt in > > > and out of certain features as they each wish to reach different levels > > > of support. > > > > > > In the second phase of the discussion, I presented a document containing > > > some of my initial thoughts on schema interfaces in TinkerPop > > > (https://gist.github.com/Cole-Greer/cf6e05f9dd30b413f3029cdecb3d4673). > > > There was quite a bit of debate over the preferred representation of > > > schemas and ways of interacting with it in TinkerPop, although most folks > > > found value in both extending gremlin's capabilities to interact with a > > > graph-representation of the schema, as well as an interface built around > > > a concise class representation of a schema. > > > > > > Our intention is to continue schema discussions in the next TinkerPop > > > gathering, which I hope will spawn some [DISCUSS] threads to fully > > > explore the concepts raised. I want to thank everyone who was able to > > > participate. I feel like I gained much greater context regarding > > > everyone's current positions on graph schema. My intention for the next > > > call is to prepare more detailed proposals for schema interfaces (and > > > perhaps even a PoC if time permits) given the discussions, as well as to > > > spend some time considering how aspects of the discussed rich type system > > > may integrate into TinkerPop without burdening providers who wish to > > > opt-out. If anyone is interested in joining, please be sure as always to > > > reach out. > > > > > > Thanks, > > > Cole > > > > > > > > > On 2026/02/26 00:29:07 Cole Greer wrote: > > > > Hi everyone, > > > > > > > > Due to lack of availability, I've rescheduled this weeks gathering to > > > > next Thursday (March 5) at 17:00 UTC. > > > > > > > > For those unaware, Josh has offered to lead a discussion on graph > > > > schema and validation. I will also be preparing materials with some > > > > thoughts on requirements for a schema API in TinkerPop. My intention is > > > > for this discussion to help spawn some [DISCUSS] threads regarding > > > > schema. > > > > > > > > As always please let me know if you're interested in joining the > > > > gatherings. > > > > > > > > Thanks, > > > > Cole > > > > > > > > On 2026/02/02 23:37:36 Yang Xia wrote: > > > > > Hi everyone, > > > > > > > > > > Here's the summary notes for the Jan 29 discussions. Thanks for > > > > > joining us! > > > > > > > > > > Attendance: > > > > > - Pieter Martin > > > > > - Joshua Shinavier > > > > > - Yang Xia > > > > > - Ken Hu > > > > > > > > > > Summary: > > > > > > > > > > The meeting focused on transaction API design for TinkerPop 4, > > > > > specifically discussing a proposal to move transactions to > > > > > GraphTraversalSource to simplify remote transaction handling. During > > > > > the discussion, various related topics emerged including latency > > > > > challenges with remote graphs, the role of TinkerPop as a language > > > > > framework, and meta-model and schema language needs. > > > > > > > > > > Transaction API Design and Related Discussion: > > > > > > > > > > A proposal was discussed for remote transactions: moving tx to > > > > > GraphTraversalSource so that g.tx().begin() starts the transaction on > > > > > the thread rather than returning another GraphTraversalSource. This > > > > > would require creating multiple GraphTraversalSource instances (using > > > > > traversal().with(DriverRemoteConnection)) for concurrent transactions > > > > > instead of reusing "g". Concerns were raised about larger interface > > > > > changes that would break users. > > > > > > > > > > The conversation touched on latency challenges when graphs sit over > > > > > remote connections (unlike TinkerGraph which is embedded with no > > > > > latency). Remote implementations, like Janus Graph, face latency > > > > > problems requiring step replacement. This led to discussion about > > > > > TinkerPop potentially positioning itself more as a language framework > > > > > rather than a protocol implementation, allowing implementers to > > > > > provide their own remote protocols. The focus could shift to grammar > > > > > and a well-structured meta-model rather than the Java reference > > > > > implementation, reducing the importance of step implementations and > > > > > using grammar to define what constitutes a TinkerGraph. > > > > > > > > > > Meta-Model and Schema Language: > > > > > > > > > > Questions were raised about validation constraints for defining a > > > > > valid graph. TinkerPop has an implicit notion of vertex, edge, and > > > > > properties, with simplicity as its strength, but lacks a formal > > > > > schema language. The need for a well-articulated meta-model and > > > > > formal description was emphasized. > > > > > > > > > > A dedicated meeting on this topic will be organized by Josh and open > > > > > to community members interested in the subject. > > > > > > > > > > Please feel free to add anything I might have missed. > > > > > > > > > > Thanks, > > > > > > > > > > Yang > > > > > > > > > > On 2026/01/27 19:55:58 Cole Greer wrote: > > > > > > Hi everyone, > > > > > > > > > > > > I've moved the meeting to Thursdays at 17:00 UTC as this appears to > > > > > > best accommodate all parties. > > > > > > > > > > > > Thanks, > > > > > > Cole > > > > > > > > > > > > On 2026/01/24 07:37:51 pieter wrote: > > > > > > > Hi, > > > > > > > > > > > > > > 17:00 UTC is fine with me. > > > > > > > > > > > > > > Thanks > > > > > > > Pieter > > > > > > > > > > > > > > On Fri, 2026-01-16 at 23:07 +0000, Cole Greer wrote: > > > > > > > > Pulling this reply from Josh into the thread: > > > > > > > > > > > > > > > > > Hi Cole, > > > > > > > > > > > > > > > > > > I am interested in attending the next one, although 8am PST is > > > > > > > > > always going to be tricky for me; that's exactly when I am > > > > > > > > > busiest > > > > > > > > > getting my kids off to school. Any possibility of 10am PST? I > > > > > > > > > realize that this would be evening in Europe. 9am PST would > > > > > > > > > also > > > > > > > > > work (I would just be a few minutes late). > > > > > > > > > > > > > > > > > > Best regards, > > > > > > > > > > > > > > > > > > Josh > > > > > > > > > > > > > > > > Hi Josh, > > > > > > > > > > > > > > > > It would be great if you could join as well. I'll leave the > > > > > > > > scheduling question open for a few days for others to weigh in, > > > > > > > > particularly Pieter and Andrii as they will be most impacted by > > > > > > > > a > > > > > > > > later start. My hope is that everyone can make it for most of > > > > > > > > the > > > > > > > > meeting starting at 17:00 UTC (09:00 PST, 18:00 CET, 19:00 > > > > > > > > SAST). If > > > > > > > > this too late for some folks, I suggest that we alternate these > > > > > > > > gatherings between 16:00 and 18:00 UTC start times. > > > > > > > > > > > > > > > > Please let me know your thoughts on the later times. > > > > > > > > > > > > > > > > Thanks, > > > > > > > > Cole > > > > > > > > > > > > > > > > On 2026/01/16 02:17:27 Cole Greer wrote: > > > > > > > > > Thank you to everyone who joined the gathering today. A > > > > > > > > > summary of > > > > > > > > > the discussions is included below. The next gathering is > > > > > > > > > currently > > > > > > > > > scheduled for Jan 29 at 16:00 UTC. Please let me know if > > > > > > > > > you're > > > > > > > > > interested in joining and if any change of schedule would be > > > > > > > > > helpful. > > > > > > > > > > > > > > > > > > Attendance: > > > > > > > > > - Cole Greer > > > > > > > > > - Pieter Martin > > > > > > > > > - Yang Xia > > > > > > > > > - Ken Hu > > > > > > > > > > > > > > > > > > Transactions: > > > > > > > > > The largest topic of conversation was transactions in > > > > > > > > > TinkerPop 4. > > > > > > > > > The discussion covered many of the differences between > > > > > > > > > embedded and > > > > > > > > > remote transactions in TP3, and how there is some intentions > > > > > > > > > to > > > > > > > > > unify this in TP4. SQLG is primarily concerned with the > > > > > > > > > embedded > > > > > > > > > use case, and the existing model works well for those > > > > > > > > > purposes. The > > > > > > > > > existing remote transaction model cannot be retained in TP4 > > > > > > > > > as its > > > > > > > > > tightly coupled to sessions, which no longer present in TP4. > > > > > > > > > The > > > > > > > > > way that embedded transactions are currently bound to threads > > > > > > > > > does > > > > > > > > > not translate naturally to GLVs such as gremlin-js. > > > > > > > > > Compatibility > > > > > > > > > with frameworks such as Spring Boot was raised as a key > > > > > > > > > requirement. There was substantial discussion around if > > > > > > > > > TinkerPop > > > > > > > > > should force the scoping of a transaction to a single thread, > > > > > > > > > if > > > > > > > > > they could be allowed to migrate between threads, or if the > > > > > > > > > API > > > > > > > > > should leave it open to implementers. The conversation model > > > > > > > > > in > > > > > > > > > JBoss Seam was raised as a potentially interesting case study > > > > > > > > > to > > > > > > > > > investigate. All parties expressed interest in continuing to > > > > > > > > > look > > > > > > > > > at other database implementations for inspiration, > > > > > > > > > considering what > > > > > > > > > requirements make sense for TinkerPop, and continuing the open > > > > > > > > > [DISCUSS] threads to build consensus on this matter. > > > > > > > > > > > > > > > > > > JPMS: > > > > > > > > > There were discussions of if TinkerPop should move to adopt > > > > > > > > > JPMS. > > > > > > > > > All parties agreed that modularization would be a good result > > > > > > > > > for > > > > > > > > > TinkerPop, there were concerns raised that lack of support > > > > > > > > > from our > > > > > > > > > current dependencies may create challenges with adoption as > > > > > > > > > well as > > > > > > > > > limiting the upside. > > > > > > > > > > > > > > > > > > Extensibility of the grammar: > > > > > > > > > Some discussion resurfaced around a past devlist post to allow > > > > > > > > > providers to extend the grammar > > > > > > > > > (https://lists.apache.org/thread/528f5od4d9jrvw9mn0b6xlhtfhvddfoc). > > > > > > > > > It was raised as a nice to have to limit the differences > > > > > > > > > between > > > > > > > > > embedded and remote usages. There was no discussion on the > > > > > > > > > mechanics of implementing such capabilities. > > > > > > > > > > > > > > > > > > Dependencies: > > > > > > > > > TinkerPop should strive to the number of dependencies (many > > > > > > > > > are > > > > > > > > > redundant or have limited usage). > > > > > > > > > > > > > > > > > > Semantics Documentation: > > > > > > > > > There was support raised for a more clear distinction between > > > > > > > > > the > > > > > > > > > gremlin language specification and the reference > > > > > > > > > implementation in > > > > > > > > > TinkerPop. The gremlin semantics docs are progressing towards > > > > > > > > > becoming a complete language specification, however that work > > > > > > > > > is > > > > > > > > > not complete and much of gremlin currently remains defined by > > > > > > > > > the > > > > > > > > > reference implementation. > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > Cole > > > > > > > > > > > > > > > > > > On 2026/01/12 23:24:48 Cole Greer wrote: > > > > > > > > > > Hi Andrii and Yang, > > > > > > > > > > > > > > > > > > > > I've tentatively scheduled this series of gatherings to > > > > > > > > > > begin > > > > > > > > > > this Thursday (Jan 15) at 16:00 UTC. I've scheduled the > > > > > > > > > > gathering > > > > > > > > > > to repeat every 2 weeks. This can of course always be > > > > > > > > > > altered > > > > > > > > > > based on availability and interest. > > > > > > > > > > > > > > > > > > > > I've sent a calendar invite to everyone who has replied > > > > > > > > > > here, as > > > > > > > > > > well as Ken and Pieter who expressed interest in Discord. > > > > > > > > > > Feel > > > > > > > > > > free to forward the invite to anyone else who is > > > > > > > > > > interested, and > > > > > > > > > > I will continue inviting anyone who asks. > > > > > > > > > > > > > > > > > > > > Others may use the following link to join the meeting as > > > > > > > > > > well: > > > > > > > > > > https://teams.microsoft.com/l/meetup-join/19%3ameeting_MTk3OTIxYzktYTU1MC00YzQzLTllM2MtMzk5YjdjMzk5MDli%40thread.v2/0?context=%7b%22Tid%22%3a%22f2267c2e-5a54-49f4-84fa-e4f2f4038a2e%22%2c%22Oid%22%3a%22f3bad5a5-c1a2-4172-b5ad-54f2ac72b2c8%22%7d > > > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > > Cole > > > > > > > > > > > > > > > > > > > > On 2026/01/12 17:34:16 Yang Xia wrote: > > > > > > > > > > > Hi Cole, > > > > > > > > > > > > > > > > > > > > > > Thanks for setting these up! I like the idea of having the > > > > > > > > > > > meeting on Teams instead of Discord, it helps to mark > > > > > > > > > > > them in > > > > > > > > > > > my calendar. > > > > > > > > > > > > > > > > > > > > > > I'm open to Wednesdays, but I do have an alternative > > > > > > > > > > > commitment > > > > > > > > > > > at 8:30am PT, so I'd prefer Tuesday or Thursday if > > > > > > > > > > > possible, in > > > > > > > > > > > case the discussions go long. > > > > > > > > > > > > > > > > > > > > > > Will the meetings start at the end of January? > > > > > > > > > > > > > > > > > > > > > > Cheers, > > > > > > > > > > > > > > > > > > > > > > Yang > > > > > > > > > > > > > > > > > > > > > > On 2026/01/10 05:29:24 Andrii Lomakin wrote: > > > > > > > > > > > > Good day, Cool. > > > > > > > > > > > > Please include me in invite. > > > > > > > > > > > > > > > > > > > > > > > > On Fri, 9 Jan 2026, 23:48 Cole Greer via dev, > > > > > > > > > > > > <[email protected]> > > > > > > > > > > > > wrote: > > > > > > > > > > > > > > > > > > > > > > > > > Hi everyone, > > > > > > > > > > > > > > > > > > > > > > > > > > I’d like to propose the establishment of recurring > > > > > > > > > > > > > meetings > > > > > > > > > > > > > to assist with > > > > > > > > > > > > > engagement and alignment across the TinkerPop > > > > > > > > > > > > > community. > > > > > > > > > > > > > I envision these meetings as a place where folks can > > > > > > > > > > > > > discuss any ideas, > > > > > > > > > > > > > concerns, or goals they may have related to > > > > > > > > > > > > > TinkerPop. All > > > > > > > > > > > > > formal proposals and development decisions will > > > > > > > > > > > > > remain in > > > > > > > > > > > > > the dev list. > > > > > > > > > > > > > These meetings would be open to anyone who’d like to > > > > > > > > > > > > > join, > > > > > > > > > > > > > and meeting notes would be taken and posted back to > > > > > > > > > > > > > the > > > > > > > > > > > > > devlist for anyone > > > > > > > > > > > > > who couldn’t attend. > > > > > > > > > > > > > > > > > > > > > > > > > > I think that approximately once every 2 weeks is a > > > > > > > > > > > > > good > > > > > > > > > > > > > cadence for these > > > > > > > > > > > > > meetings. As most of the folks I see active here are > > > > > > > > > > > > > located > > > > > > > > > > > > > in Europe and North America, I think a time of 16:00 > > > > > > > > > > > > > UTC > > > > > > > > > > > > > (17:00 CET, 08:00 > > > > > > > > > > > > > PST) is a good compromise to start with. I’m > > > > > > > > > > > > > certainly open > > > > > > > > > > > > > to other meetings times as folks express their > > > > > > > > > > > > > availability. If these > > > > > > > > > > > > > meetings draw interest from folks in Asia, Europe, > > > > > > > > > > > > > and the > > > > > > > > > > > > > Americas, > > > > > > > > > > > > > then I would suggest we adopt a rotation of times > > > > > > > > > > > > > such that > > > > > > > > > > > > > there are some > > > > > > > > > > > > > meetings available at a reasonable time to everyone > > > > > > > > > > > > > who is > > > > > > > > > > > > > interested. > > > > > > > > > > > > > > > > > > > > > > > > > > I’d suggest scheduling the meetings every 2nd > > > > > > > > > > > > > Wednesday, > > > > > > > > > > > > > although Tuesdays > > > > > > > > > > > > > or Thursdays are also good if anyone expresses a > > > > > > > > > > > > > preference. > > > > > > > > > > > > > > > > > > > > > > > > > > If there is sufficient interest, I would expect these > > > > > > > > > > > > > new > > > > > > > > > > > > > meetings to take > > > > > > > > > > > > > the place of our current gathering on Discord. I > > > > > > > > > > > > > would send > > > > > > > > > > > > > invites to a Teams meeting to anyone who is > > > > > > > > > > > > > interested, as > > > > > > > > > > > > > well as making > > > > > > > > > > > > > a meeting link publicly available. Anyone will be > > > > > > > > > > > > > able to > > > > > > > > > > > > > join > > > > > > > > > > > > > without creating an account, as well as optionally > > > > > > > > > > > > > via a > > > > > > > > > > > > > browser. > > > > > > > > > > > > > > > > > > > > > > > > > > Please let me know if you are interested in such > > > > > > > > > > > > > meetings > > > > > > > > > > > > > and if you have > > > > > > > > > > > > > any preferences on scheduling. > > > > > > > > > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > > > > > Cole > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > >
