The problem I think is that you tie the Lucene and Solr releases together, and expect that whenever a new Lucene release is out, with features that are considered great for the *Lucene* users, the matching Solr release will have a number of cool new features too.
We've decided a long time ago to release Lucene and Solr together. But I don't remember that we decided that each Lucene/Solr release needs to include great/cool features on both of the releases. Yes, if Lucene's code base is hardened, enough to warrant a dot release, and our policy says this means a dot Solr release as well, then Solr users are advised to upgrade (just like ElasticSearch users, when they will issue their dot(.dot) release or whatever. Lucene is the base of Solr, and I see a justification for e.g. a Solr 4.10 even if the only lines of code that were changed are in the Lucene binaries that Solr 4.10 will package. That, to me, is completely orthogonal to how often Solr releases cool and great new features. If you feel this hasn't been the case since Solr 4.1 (and I don't think this is the case), then this is something serious we should discuss in the context of "how come Solr didn't release something major in last 2 years?". But as I said, this certainly isn't the case. If you're looking for a justification to upgrade to Solr 5.0, then I will try this: "because the Lucene/Solr community has decided to release version 5.0, which improves code stability, gets rid of back-compat baggage which resulted in bugs, index corruptions and prevented code improvements and enhancements; and because Solr and Lucene are released together - this is mainly a release that will get your Solr on par w/ the latest Lucene binaries, and will guarantee that your indexes will get supported by next Lucene/Solr major version - 6.0; As usual, Lucene and Solr will continue to release enhancements and cool features throughout the dot releases". Point is (at least on my part) - *our* (this community's) major releases are mostly about index backwards compatibility support. We also note big API improvements/changes, though we that's not always the case (see 4.2 w/ the DocValues changes, 4.1 and then again 4.7, or 4.8 with facet APIs overhauling) - depends if we marked a feature experimental or not. But to the user who e.g. used DocValues in 4.1, and had to migrate to 4.2, I don't think it would matter to him much if we said 4.2 is actually 5.0 -- it's the same amount of work to him. So that's why I conclude that our major releases are mostly about index back-compat support. If it happens to be that they also come w/ great and cool new features, that's a bonus IMO. For instance, and I'm only speculating, I think it's just luck (that's my speculation) that both Lucene *and* Solr 4.0 had major features (Lucene: Codecs, Solr: SolrCloud). The two are completely unrelated and it could very well be that we would release Lucene 4.0 (and Solr to accompany it) before SolrCloud is ready. And when that's ready, it would follow with either a 4.1 or 5.0 release (both would be fine by me). But I think that attempting to look at Lucene/Solr major releases as if they carry some sort of gospel, is wrong. A 5.1 release can easily be bigger than e.g. the previous 4.0 (at least in # features and lines of code changes). We don't need to wait for 6.0 with this. Our users should upgrade to the next major release, I would say, ALWAYS. That's the only way they can guarantee their indexes continued to be supported. Even if that major release has 0 (note, ZERO) # changes, they should still upgrade. They will need to anyway, when the next-next major release is out (or they reindex). Shai On Mon, Oct 13, 2014 at 7:11 PM, Jack Krupansky <[email protected]> wrote: > (And then there are people like me and my "Solr 4.x Deep Dive" book - I > guess I'll finally have to get back to catching up, and have a great excuse > for a new/revised title.) > > -- Jack Krupansky > > -----Original Message----- From: Alexandre Rafalovitch > Sent: Monday, October 13, 2014 10:49 AM > To: [email protected] > Subject: Re: Is there going to be Lucene/Solr 4.11? > > On 13 October 2014 10:37, Jack Krupansky <[email protected]> wrote: > >> Now... I'll have to admit that maybe there might be clarity among the >> Lucene >> dev/user crowd, but mostly I'm referring to the Solr user crowd, who >> aren't >> up on Lucene internals. >> > I do admit having troubles envisioning the Solr 5 introduction book > start with information I see so far: > "Solr 5 is a great new release. _Users_ will benefit from using the > new amazing features X, Y, and Z that take Solr beyond already great > capabilities in version 4 and now allow <scenario 1>, <scenario 2>, > and <scenario 3>". > > I am not saying there is nothing to slot into that phrase above. But > in all the discussions, it seems unclear what the actual user benefits > will be. The index-corruption is an internal - if very important issue > - but not something we can phrase user benefits around. "Increased > stability" is a user-level term I guess, but it is semi-negative as it > automatically implies previous _general_ instability. > > This is an ongoing problem with Solr - changes that occur strictly in >> Lucene >> with no source changes in Solr, never get reported as improvements to >> Solr - >> and it is nigh impossible for a Solr user to examine a list of Lucene >> changes and determine how they may or may not impact the use of Solr >> features. >> > > I confirm this perception. I remember looking at one of the recent > Solr release notes and thinking: "hmm, not much here, why release" to > then click-through to the Lucene notes and seeing the significant > number of important items. Most of the people do not click-through > though and those who do, do not necessarily understand which of the > issues do have impact on Solr. > > I guess what I am saying is that it would be nice for somebody to > figure out the Solr 5 user story at some point _before_ we actually > shipped it. And the earlier, the better. > > Regards, > Alex. > P.s. If somebody wants to discuss this user-level perception at Solr > Revolution, I am happy to shout the first round of drinks. :-) Or, > hash it out virtually in the popularizers community I setup: > https://www.linkedin.com/groups?gid=6713853 > > > Personal: http://www.outerthoughts.com/ and @arafalov > Solr resources and newsletter: http://www.solr-start.com/ and @solrstart > Solr popularizers community: https://www.linkedin.com/groups?gid=6713853 > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
