Hi all, +1 for option (a).
The previous comments are absolutely valid: Models won’t be released as often as the OpenNLP core artifacts. With this in mind, option (b) is somewhat artificial and could even confuse the community, in case there is no „most recent“ model release that matches the tools artifact. If documented accordingly, option (a) should be clear to devs and users. [at] Jeff: Thanks for opening LEGAL-676 Best Martin -- > Am 28.05.2024 um 22:19 schrieb Jeff Zemerick <[email protected]>: > > I favor option (a) because we likely won't release models as frequently but > we will have to keep track of what's compatible with what. > > There is a manifest file inside the model files and it contains the version > number of OpenNLP that trained the model. It's used to check if the version > of OpenNLP loading the model can actually use it. When we released OpenNLP > 2.0 that check had to be changed because it was only looking for versions > starting with 1.x. So just something to keep in mind that we might need to > change that check. So far all models are compatible with all versions. > > Also, I opened a ticket with legal to see about bringing the SourceForge > models in to make it easier to distribute them. > https://issues.apache.org/jira/browse/LEGAL-676 > > Thanks, > Jeff > > On Tue, May 28, 2024 at 4:03 AM Atita Arora <[email protected]> wrote: > >> hi, Thanks for initiating this discussion regarding the future version >> scheme for our OpenNLP Maven module distribution. >> Personally, I lean towards option (a) as it establishes a fresh starting >> point for our Maven module distribution. >> However, I'm open to hearing others' thoughts and considerations. >> >> Best, >> Atita >> >> On Tue, May 28, 2024 at 8:31 AM Richard Zowalla <[email protected]> wrote: >> >>> Hi, >>> >>> From my point of view, we are not releasing models as often as we release >>> opennlp-tools. >>> And pushing a new version without having actual changes in supplied >>> models, looks somehow odd to me. >>> So personally, I would be in favor for having a separate versioning >> scheme >>> for it. >>> >>> Gruß >>> Richard >>> >>> >>>> Am 27.05.2024 um 21:56 schrieb Bruno Kinoshita < >> [email protected] >>>> : >>>> >>>> Hi! >>>> >>>> Thanks for starting this! I do not have a preference for a, b, or c, as >>>> long as it's explained/documented to users, I think they will be happy >> to >>>> be able to see the model in their dependency tree. >>>> >>>> Thank you! >>>> >>>> On Mon, 27 May 2024 at 16:15, Richard Zowalla <[email protected]> wrote: >>>> >>>>> Hi all, >>>>> >>>>> with [1] we will make a step forward in order to distrubute OpenNLP >>> models >>>>> via Maven Central. >>>>> While we still need to integrate all other models, Martin (W.) raised >>> the >>>>> valid question about the future version scheme of our opennlp maven >>> module >>>>> distribution. >>>>> >>>>> Do we want to >>>>> >>>>> - (a) Start with 1.0 >>>>> - (b) Align with the version of the opennlp-tools >>>>> - (c) Something else >>>>> >>>>> Any ideas, thoughts, comments? >>>>> >>>>> Gruß >>>>> Richard >>>>> >>>>> [1] https://github.com/apache/opennlp-models/pull/1 >>>>> >>> >>> >>
