I am pleased to finally be able to announce that we will
begin preliminary testing of Ingrid software. We are
looking for a small group of volunteers.
What Ingrid Is:
For those of you who haven't followed Ingrid, it is a
technique for whole-web searching that doesn't require
a central search database. Instead, it automatically
generates links between similar web resources, resulting
in an infrastructure that can be (we hope) efficiently searched
robot-style (at search time). While it is certain to have
disadvantages to central search databases, one advantage
we hope it has is that it pushes the control over searching
to the edges (publishers and navigators). Anyway, please
see http://rodem.slab.ntt.jp:8080/home/index-e.html for
more info (please don't imagine that the quality of our
home-page is a reflection on the quality of our software :-)
What We Have:
We have early versions of software that:
1. Publisher: Finds (using a robot) and processes text resources
(isolates words, selects high-weight terms, etc.) in English and
Japanese (please contact us if you want to add your favorite
language...character set is not a restriction).
2. Server: Inserts the processed text into a global linked
infrastructure, and then answers queries about that text and text
that it is linked to.
3. Navigator: Allows users to search/navigate that infrastructure
using a robot/GUI navigator running on their own workstation. This
navigator has some nice relevance feedback features you don't
find on current web search engines (that I'm familiar with).
The navigator can also be coupled with your browser.
The Publisher is in Perl. The Server is in C, and the Navigator is
in C and uses Motif. We are running on Sun OS4 and Solaris. All
interactions take place using UDP packets with a high port number.
(This software, by the way, will all be freely available.)
What We Want To Do:
Starting around the end of this month, we want to start building the
infrastructure. The primary purposes are:
1. to try to find out how well it scales,
2. to flush out bugs, installation procedures, etc.,
3. to learn what kind of features we should add to the navigator.
Who We Are Looking For:
On the publishing side, we need voluteers that:
1. Are willing to spend a bit of time with this. While we are trying
to make this all as simple as possible, you will have to spend
some time at installation, check to see that things are working
properly, and so on. Also, we expect there to be frequent
re-installs early on, for bug fixes etc.
2. Have a machine that is continuously internet-accessible and has
some spare memory/CPU/bandwidth. (Just how much we don't know...that
is part of what we want to learn. It should, however, be negligible
compared to running a web server.)
3. Ideally (though not absolutely necessary) you should have a set of
resources that you yourself would like to be able to search (and
that you would like others to be able to search). In particular, we
are able to loosely couple Ingrid with other local search engines in
that our search result not only finds resources for you, but points
you to search engines that contain those resources so that you can do
additional searching. Also, if you've already installed Harvest, our
software can use your Harvest configuration files, so you don't have
to do it again.
I would say that our ideal first "customer" is someone that has already
installed multiple Harvest (or similar) databases, and would like to
see those databases coupled in some nice way. It is ideal because
such a person gets some immediate personal benefit from using Ingrid.
What You Should Do:
Please send me mail (francis@slab.ntt.jp) if you are interested in
participating. (Do so even if you want to postpone your participation
till we have something more solid.) We can then discuss particulars.
Thanks,
PF