Hi, I am talking about optimizations/features that are not part of TinkerPop's api.
I recently tried BlazeGraph after all the talk about how fast it is. Turns out it is slow, very slow. https://github.com/blazegraph/tinkerpop3/issues/3 Turns out they have a custom bulk load api, how would a client via GremlinServer use that? Sqlg has similar optimizations/features outside TinkerPop. Titan has a whole world of types and indexes and and and. I reckon most databases will have such "features" If a client only accesses a graph via GremlinServer how would they use such optimizations? Perhaps via some plugin framework? Thanks Pieter On 23/03/2016 12:33, Stephen Mallette wrote: > Dylan, thanks for your thoughts here. The Gremlin Server docs are really > just for reference at this point. As you've pointed out, we could do more > to improve people's understanding of that part of the stack. I've been > hesitant in writing more documentation in this area as I've been waiting > for more of the stack to settle and better implementations patterns to > emerge. I think that started with RemoteGraph which we will have in 3.2.x > as that's opened up a lot of new ideas both on and off the JVM as we > consider ways to stop doing the classic SQL-ish way of passing long strings > to a database server. I don't know that we need a Gremlin Server specific > site - but maybe i'm not sure of what you have in mind. I've been thinking > that the time is coming where we will want to have a solid Gremlin Server > tutorial though (perhaps several with different connectivity patterns for > different languages). ?? > > Pieter, you don't lose any backend optimizations with Gremlin Server. The > same Traversal Strategies that would fire in embedded mode would fire in > Gremlin Server. > > On Tue, Mar 22, 2016 at 3:14 PM, pieter-gmail <pieter.mar...@gmail.com> > wrote: > >> Hi, >> >> What I would add to the gist is the latency penalty. As a java guy and >> db guy its the bane off my live, OO devolves into query strings because >> of latency and eventually one feels like one would be better of with >> 60's tech writing stored procedures and PLSQL. >> >> Anther one is the fact that dbs compete in the space where they do not >> follow the specification. Does Gremlin Server make one loose any >> particular backends optimizations? >> >> I have never used Gremlin Server so perhaps my concerns are moot, but if >> they are on a website like you propose I'd read it. >> >> Cheers >> Pieter >> >> On 22/03/2016 20:31, Dylan Millikin wrote: >>> Hey guys, >>> >>> Coming from a non-java background I've been thinking about what steps >> could >>> be taken to improve the project's reach in these communities. As time has >>> gone by I've added some thoughts to the following gist (they are just >>> thoughts at this point, some may be unrelated or simply not possible) : >>> https://gist.github.com/PommeVerte/2fc463be9c418b465456 >>> >>> The one I want to talk about in this email is about the benefits of >> having >>> a Gremlin server dedicated space on the website (ideally with it's own >> URL >>> for clarity and SEO purposes, though I'm unsure what Apache guidelines >> are >>> here). >>> >>> Right now if I'm new to graph *databases* (emphasis) and am looking to >> get >>> started, the TinkerPop website is not inviting. There's a lot of >>> information about the stack (purposfuly so) and I'm instantly lost in a >>> bunch of information I can't relate to. So I just go to Neo4J and learn >>> cypher ;p >>> >>> I understand that Gremlin server is a commodity for implementors and not >> an >>> actual database. But I think there's value in promoting it as one. Typing >>> "graph database" in google should bring Gremlin Server up, if only to >> make >>> people aware that they have the option of using it with the DB of their >>> choice. >>> >>> I wanted to reach out and hear what your thoughts on the subject were. >> Also >>> perhaps this is within the scope of "tinkerland" and is already being >> taken >>> care of. I wouldn't know at this stage. >>> >>
