Cluster update

[Gerald Combs]

On 6 Dec 2002, david nicol wrote:

> On Tue, 2002-12-03 at 17:16, Gerald Combs wrote:
> > How about doing regression/stress testing of open source software?  We
> > could run a bunch of packages under Hal's memcheck utility, and throw
> > random data at them.  Sort of a Q/A department for Freshmeat and
> > SourceForge.
> 
> 
> We could give it an old-west kind of name and have a drawing
> of westport in the nineteeth century as its flag.  "Does your
> package pass the gunslinger test?"

Well, Red Hat already took the name "Rawhide."  "Regressionsmoke" isn't
catchy at all.  How about "How the source was won?"

Back when I worked at an ISP that shall remain nameless, I had an idea for
a project that would be perfect for a massive cluster of cheap machines:

Access blocking software such as the kind used in schools, libraries and
corporations works by having a database somewhere containing URLs that
have been assiged different categories.  If you request a URL that's been
listed in a "block" category, the software keeps you from accessing that
URL.

No matter what company makes the software, these databases have two common
features: they are always Proprietary and Secret (meaning you can't see
the lists of URLs), and entries are added based on political and moral
views that may or may not match your own.

What if we made a block list database completely open, and allowed anyone
to create and maintain their own URL categories and lists?  You could then
pick and choose who you trust to censor your web experience.  For example,
if Joe, Louise, and the National Whig Party all had their own block lists
you could configure your blocking software like so:

   block_from: joe OR louise AND NOT whig_party

A cluster of machines would be needed to store and manipulate a database
of any useful size.