Re: A bit of bomb throwing....

From: Rickard Öberg <>
Date: Tue, 17 Jan 2006 20:42:06 +0100

Zack Rosen wrote:
> That said, from personal experience, I don't see a way RDF object
> stores will be pervasively implemented in web-technology such as
> content management systems any time soon.

Perhaps I should mention that the product I work on (and am, in fact,
chief architect of) is a CMS/portal. And that we're the second largest
vendor in Sweden of said types of products.

> If you were to propose this
> to the Drupal community you would probably get laughed off the
> mailinglist. The problem is the burden of proof as to the benefits of
> adopting these technologies and concepts still remain upon the research
> community and RDF object stores remain relatively unproven. In the
> communities I work with this is exacerbated by the fact that
> experiments with RDF object stores have been nothing short of
> disastrous. One multi-$100K project I know of employed an RDF store in
> their CMS/Communtiy application only to see it choke under light web
> traffic loads. The website never launched and the project was scrapped.

As already mentioned I did a spike test of using Sesame as our backend
store today, and as already mentioned performance is one of the key
factors for me, and as already mentioned I did not see a significant
difference between Sesame and our current persistence store.

What I haven't already mentioned is that in my experience the backend
store has a very small effect on overall performance of a CMS. A strong
architecture and LOTS AND LOTS of caching, now that's where you get
serious performance. As a funny note we just broke the performance
record of one Major Government branch when they redid their website
using our stuff. The funny part is that our CMS is entirely based on
dynamic generation of pages and the old one used Apache and static
content. So again, performance in a CMS basically comes down to: "how
little can you touch the disk?".

As far as I can tell, the RDF store will be great for *finding* content,
but once I have located the id I will use more standard techniques to
actually access and use it, and for the most part I hope to have it in
my caches already anyway. That approach has worked very well so far anyway.

> My point is not to spread doom and gloom though. I just think we need
> to objectively think about what it will take to get these tools widely
> adopted.

I agree, objectivity is very important when assessing these kinds of things.

Received on Tue Jan 17 2006 - 19:41:46 EST

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT