Jason Plurad <plu...@...>
Hi Nikolai,
If you're happy with Titan 1.0 and you are comfortable putting it into production and supporting it by yourself, I'd say go for it. There are folks out there with Titan 1.0 in production. They are likely maintaining private forks with any fixes they needed and applied themselves.
> I would like to see the completion of Titan 1.1 work as JanusGraph release and then future development.
Are there specific functions in Titan 1.1 you were waiting on? If so, I'd suggest opening up issues on the JanusGraph tracker. We didn't copy over all of the open issues from the Titan issue tracker. As far as I can recall, the Aurelius crew didn't publish any specific roadmap of features for Titan 1.1. DataStax owns Titan, so it would be up to them to publish a Titan 1.1 release, however they made it abundantly clear they were not interested in doing so. In addition to an open community with open governance, a major goal of the JanusGraph project was to consolidate private forks from other in the community so we can easily migrate existing Titan users and continue forward with a single codebase. Misha mentioned that JanusGraph is forked directly from the latest Titan 1.1 development branch, so JanusGraph is essentially a continuation of Titan. As a consumer, you benefit from code fixes and features provided by others in the community. And there have been many fixes put in already! Development priorities have been mostly around upgrading dependencies in that had gone stale because Titan had been out of development for a long time. Pretty much all of the core dependencies (TinkerPop, Cassandra, HBase, BerkeleyJE, Elasticsearch, Solr, Lucene) have been updated to work with recent versions. There is work going on to improve the test suite to ensure we have a solid platform before introducing significant breaking changes.
> There are some bugfixes but surely there may be new issues introduced and not yet found :)
True of any software project, open source or not.
> What makes the choice even more difficult is that I understand that it is my (as everyone's else) job to actually find, report and, if possible, fix these issues ;)
That's ideal state for any open source software project. But even if you can't fix it yourself, reporting it is a solid contribution regardless. If the bug is significant enough, others are likely running into it and would be motivated to help fix it. If you stick with Titan, there's little point in reporting it publicly since it doesn't appear that DataStax will fix that codebase publicly anymore. The burden would be all on you.
-- Jason
On Tuesday, July 4, 2017 at 11:12:24 PM UTC-4, Nikolai Grigoriev wrote: Hi Misha,
I think I misused the word "stable" - it is all relative in software in general and large open-source projects in particular. And, of course, it is not all about a nice number like 1.0.
I knew that JanusGraph has been forked from relatively stable branch of Titian. I am looking at it from pragmatic point of view of someone who did not diff Titan 1.0 and latest JanusGraph release. There are some bugfixes but surely there may be new issues introduced and not yet found :) What makes the choice even more difficult is that I understand that it is my (as everyone's else) job to actually find, report and, if possible, fix these issues ;)
Anyway, I am trying to decide now: if I was happy about the state of Titan 1.0.0 and I would go to production with it, say, 6 months ago - should I consider JanusGraph instead of it now or not? I have not seen any kind of release strategy for JanusGraph yet. I think it would be a good idea to openly state the priorities for JanusGraph development. Is it more about converging on a relatively bug-free release with critical and low-risk bugfixes first and then going forward with breaking changes after that (or in parallel)? Or it is something else?
I think, personally, as Titan/JanusGraph user I would like to see the completion of Titan 1.1 work as JanusGraph release and then future development.
On 4 July 2017 at 18:27, Misha Brukman wrote: Hi Nikolai,
Can you please define what you mean by "stable release"? Are you looking for a 1.0 release? That's probably not happening for a while — the next release's version will be 0.2.0.
That said, you're running Titan 1.0 — note that JanusGraph 0.1.0 forked from the (unreleased) Titan 1.1 branch, and has upgraded a number of dependencies and added a number of changes and bug fixes, so it's already well ahead of Titan 1.0, rather than a new from-scratch project. While it's true that in most projects, 0.x versions are considered "development" and not for production use cases, in this particular case of JanusGraph, the version number is not nearly as important given the project's heritage.
Here's an incomplete list of companies who are already running JanusGraph in production (i.e., companies that have publicly stated that they are and agreed to be listed): and more are coming soon.
Hope this helps, Misha On Tue, Jul 4, 2017 at 11:11 AM, ngrigoriev wrote: We are going with titan-1.0.0 for now as (correct me if I am wrong) JG does not have a stable release yet.
-- Nikolai Grigoriev
(514) 772-5178
|
