- From: Dave Beckett <dave@dajobe.org>
- Date: Mon, 08 Sep 2008 09:31:27 -0700
- To: Aldo Bucchi <aldo.bucchi@gmail.com>
- CC: Kingsley Idehen <kidehen@openlinksw.com>, "public-lod@w3.org" <public-lod@w3.org>
Aldo Bucchi wrote: > Hi All, > > Sorry if everyone is asking the same thing. > Triplr went missing this weekend ( at least to us folks down in LatAm ). This was the weekend I was changing my email, DNS and networking config. If you'd have pinged me, I'd have fixed it quickly. It's working right now. Dave > I believe triplr to be an *incredibly* smart and useful service ( > especially callback wrapped JSON ). It will probably do more in behalf > of the semantic web, outreach-wise, than most other efforts out there > ( once it finds its way into small widgets and tutorials with > practical applications ). > > But, with great power comes great responsibility. > > This outage crippled several demos I was planting out in the wild, and > the perceive damage is rather huge. "The Semantic Web is Down??" I got > from a couple of High level execs. Of course there are no SLAs and no > commitment on triplr's behalf, but the ( naive and irresponsible ) > expectation is that the service doesn't break. > > I agree this is the developer's mistake but then again, the value > proposition was to make it easier for them to play around with SW > data. > > This brings up the some FUD from the consumer's perspective. > And then the same can be argued from the service provider's POV: > > If someone builds a scutter on top of Triplr, for example, and tries > to index the entire foaf-space, numbers can get big. > > What are the rules? > Should they enforce some sort of limits? > Who's paying for infrastructure? > ( Dave? ) > > Etc... > > I am now changing all these small demos to fallback to an alternative > proxy ( my own, running powered by virtuoso running on the cloud ) > and/or OpenLink hosted Demo endpoints. > > My question is: > > Has anyone given deeper thought to this issue? > > For example. How about a Pay per use proxy service ( akin to Amazon S3 > ) using signed headers that could be financially viable and provide a > SLA? > ( and more complex services, like delegated auth and session management ). > > Or perhaps a directory of open (and/or payed) rdf-izer services that > shared the same REST interface and could be swapped. Triplr, OpenLink > demo, etc. > > ( of course the latter brings up trust issues and whatnot, but we can > factor those in as well ). > > The point is that RDFization of heterogeneous resources will remain a > complex ( and evolving ) topic and a reliable simplification layer is > definitely beneficial. > > Thanks, > A >
Received on Monday, 8 September 2008 16:32:15 UTC