I'd be interested in joining such a discussion.

On 2025/11/20 19:39:58 Ken Hu wrote:
> In TinkerPop 4.x, we're going to have more options since the server is
> likely to host more endpoints (e.g. status). This opens up new
> possibilities with how the GLVs can interact with the server and in
> particular with different providers/vendors. I think we should have an open
> discussion on these topics that you have brought up on the dev list
> recently. Maybe we can schedule an open meeting for the first week of Dec
> (to avoid the Thanksgiving holiday)?
> 
> If anyone is interested in discussing some of these items then please reply
> to this thread. We can decide on a time that works for everyone in several
> days after anyone that is interested gets a chance to say so.
> 
> On Thu, Nov 20, 2025 at 9:05 AM Andrii Lomakin
> <[email protected]> wrote:
> 
> > Good day.
> > Let me provide one more argument.
> >
> > Not so long I read the book 'differentiate or  die' that is important point
> > for vendors as with tool that promotes unification by default they can't
> > differentiate themselves so efficiently  and prefer tools that promotes
> > differentiation.
> >
> > I think that is valuable point.
> >
> > On Thu, 20 Nov 2025, 14:53 Andrii Lomakin, <[email protected]>
> > wrote:
> >
> > > Good day.
> > > I understand that it contradicts current 4.x goal.
> > >
> > > To decide I propose to check how many vendors can practically work
> > without
> > > their dependencies added , I also propose to take into account impact of
> > > each vendor on infrastructure. I have a feeling that feature rich vendors
> > > can't work without their dependencies added.
> > >
> > >
> > > As one more argument JDBC users work in this way all the time and don't
> > > see any issues with this approach.
> > >
> > > On Wed, 19 Nov 2025, 19:49 Andrii Lomakin, <[email protected]
> > >
> > > wrote:
> > >
> > >> Good day,
> > >>
> > >> As Ken Hu correctly noted in a separate thread, the fact that users
> > >> sometimes ignore vendor libraries is leading to confusion.
> > >>
> > >> I propose changing how users obtain a RemoteGraphTraversal instance.
> > >> Instead of allowing direct creation of the instance, I suggest using a
> > >> method similar to RemoteGraphTraversalManager.connect(url, name,
> > password).
> > >> This new approach would enforce registration of the provider library by
> > >> throwing an exception if it is missing.
> > >>
> > >> I recognize that this proposal may be controversial, but I believe it is
> > >> worth considering as a solution to the long-lasting issue.
> > >>
> > >> Looking forward to reading your opinions.
> > >> YouTrackDB development lead,
> > >> Andrii Lomakin.
> > >>
> > >
> >
> 

Reply via email to