Clojurians Log v2

A paper on a project using grafter just got accepted in the semantic web journal: http://www.semantic-web-journal.net/content/datagraft-one-stop-shop-open-data-management-0

might be of interest to some

cool! 🙂

quoll: who are you friends with?

Tom Heath

ahh cool - I know Tom quite well 🙂

we interviewed at Talis together

ahh cool old school

I saw him at a couple of Semweb conferences, but I don’t get sent to those any more

He's now at Arup

where are you now quoll?

Cisco

trying to convince them to use RDF 🙂

but they’re letting me build a rules engine, so it’s all good

nice 🙂 Rule engines are fun

using clojure?

of course!

yay!

core.logic?

My third rules engine. My first was in Java. My second was Clojure and closed source. This one is open source

or rete or something else?

no, it’s datalog not prolog

ahh cool

rete based

https://github.com/threatgrid/naga

I’m building it specifically to be backable by either Datomic or a SPARQL store

I used to work for a startup that made a rule engine... As far as I know at the time - it was the first engine that supported defeasible inferencing with priorities

and the first thing I’m asked to? My own (i.e. not based on a third party library) database

defeasible? Ooooh!

Yeah 🙂 it was fun

Naga has priorities, but I never considered defeasibility

implemented as a meta-interpreter in prolog

Naga is entirely its own thing. But the storage is abstracted, so it can use one of several

The thing about using Prolog (or any backtracking rule system, like Drools) is that it can’t handle databases efficiently.

It was academically on very sound foundations - built upon work in defeasible logics by Henry Prakken at Utrecht and Chris Reed at Dundee

Very nice

yeah - sadly the company couldn't really find a market for it -- far too technology driven.

same with my previous company

I always meant to build it again, since I’ve heard people crying out for it on Datomic

Jena has a capable rule system, but Jena never scaled well

yeah something built on datalog/datomic would have nice practical properties

so I hear

Coincidentally, Datomic is built along very similar lines to Mulgara (a project I run… though I’ve let it run down in recent years).

we used to use Fuseki as a triple store - but it couldn't cope with the data volumes... plus Jena's API's aren't that clean/consistent in my experience.

I’ve been rebuilding the Mulgara indexes in Clojure. They have nearly identical properties to Datomic but load data faster, so I’m hoping to integrate it into Naga

cool... I've stumbled across Mulgara before - but know nothing about it

I was one of the original developers on it (we started in about 2000)

sweet

where are you based?

central VA

I telecommute

forgive me VA?

sorry… it’s the common abbreviation for Virginia, USA

ok I suspected that

but I'm British so wasn't sure

I’m Australian, so I can appreciate not knowing

are you based in the UK?

yeah... I work in Manchester at swirrl http://www.swirrl.com/ live in Yorkshire...

I’ve been a bit apart from SemWeb business recently. How is the market? That was the thing that everyone always struggled with. Great technology, but unable to get customers to invest in it

our market is good but small. We target government data publishers - I don't consider us to be a SemWeb business though... We tend to align more with Linked Data than SemWeb if you see the difference... And to be honest even though we're pretty much 100% Linked Data from a technology perspective, we tend to de-emphasise Linked Data (though I think we still emphasise it too much) and focus on the real problems and benefits.

my feelings are that aside from the true believers - if you talk about URI's, dereferencing and triples (let alone SemWeb/Ontologies etc..) in the worst case you've probably lost a sale... and in the best case you've found yourself in the free education market - and that's not a business most people want to be in 🙂

I agree that selling yourself as a SemanticWeb company isn’t a good idea. Nobody needs semantic web. They need to manage data. They need to connect data from different sources, in intelligent ways. They need to be able to analyze and find connections in that data. Any mentioned of how that gets done is a mistake IMO

other companies don’t do that. Hardware companies don’t advertise that they build chips with advanced lithographic techniques for the implementation of even smaller features in CMOS. That info is usually available, but not part of the sales pitch

Too often, I saw SemWeb companies failing because they were trying to sell the technology to people who didn’t understand that they could implement this task, or save money over here, if they took it on and built something with it. But that was due to the maturity of the technology and community. I’ve been hoping that has been improving over time

+1