On Tue, 04 Apr 2017 21:33:48 +0200, Andy Seaborne <a...@apache.org> wrote:

i am sure to run an
official Fuseki-'Reference' public endpoint is a very harmless and for
everyone comprehensible suggestion...

Open source, at Apache at least, is about community.

The code is available and anyone can put use it to put up a server. That is how the costs get shared. There is no distinction of "official".

People-capacity is the biggest limitation to everything for open source projects.


The project has "sparql.org" (it's on Apache hardware - it's a little VM) which I set up - the management and allocation of project VMs is by the Apache infrastructure team who have it highly automated. They automate everything - economies of scale.


Even keeping "sparql.org" up vaguely 24x7 isn't trivial. Most of the time, nothing needs doing.

But there are short bursts of activity needed to fix things. Last week it went down - I was travelling - and I get twitter reports. Turns out that the VM is hosed in weird and wonder VM/memory ways and it needs Apache Infrastructure to reboot - that problem took time to track down.

sparql.org gets all sorts of web-rubbish coming its way as well.

That includes DOS attacks and people hitting it with repeated queries ... in parallel.

It's a small VM -- we are lucky enough to get free.

The Apache Software Foundation runs almost entirely on volunteer effort - there are a few (small single digits) employees keeping the core (email, archives, SVN, git) infrastructure up and running. The rest of the infrastructure is kept up and running by volunteers (e.g. the build server).

It would be nice to put up a more interesting data set on sparql.org and to take time refresh its UI (it's the old Fuseki1) with a nice SPARQL query editor/form.

Something small - 1-2 M triples to run in memory.

And it needs people-time.

        Andy


All this means in practice also:

The most important app-case of Fuseki (running as a public endpoint) cannot be demonstrated by its developers in a real-world-scenario with a well-known database like dbpedia, because this costs too much and there is nobody to finance this.

But small steps with sparql.org as Fuseki-public-endpoint are possible...

Thanks for the clear and detailed report, how the situation at moment is...

baran

--
Using Opera's mail client: http://www.opera.com/mail/

Reply via email to