@Shawn: bq: Do you anticipate that we would eventually fold the CoreAdmin functionality right into the Collections API, or would it simply remain an invisible API that an expert could use if they really wanted to?
Frankly I haven't thought that far ahead. I guess (and I have no strong preferences at this point) that it's a waste to move stuff all around for equivalent functionality. Once we move the CoreAdmin to an "implementation detail", we can then do whatever makes sense. The path of least resistance would be to continue to use it as-is, make it invisible/expert and migrate things from the CoreAdmin to Collections APIs it makes our lives easier. Architecturally having the Collections API deal with, well, collections and the CoreAdmin API to realize each of the necessary actions on each instance which is the CoreAdmin API makes sense. Practically I don't want to move lots of things around for no added functionality as that's a fine way to introduce instability for no gain. Deprecating support for the CoreAdmin API would allow us to "do the right thing" as necessary as far as the Collections API, and not be hobbled by back-compat. I think I hear an echo of the Lucene guys' pain here. As far as SOLR-7316 is concerned, I don't really care whether we go with that one or not. My reactions to Hasse's tone is a personal problem, entirely unrelated to the validity of his points. I'll get over it ;). @Yonik: bq: +1, this is the position I've advocated in the past as well. Damnit! About the third legacy project I was migrating to Lucene (around 2007/2008?), I thought "gee, I seem to be doing the same kind of thing over and over again making analysis chains, I wonder if this could be done by some kind of configuration?". Must you _always_ be a step or two ahead of me ;). Erick On Sat, Mar 28, 2015 at 6:16 PM, Yonik Seeley <[email protected]> wrote: > On Sat, Mar 28, 2015 at 2:27 PM, Erick Erickson <[email protected]> > wrote: >> Fold any functionality we still want >> to support at a user level into the collections API. I mean a core on >> a machine is really just a single-node collection sans Zookeeper, >> right? > > +1, this is the position I've advocated in the past as well. > > -Yonik > > --------------------------------------------------------------------- > 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]
