There are four parts to this video. One of the most interesting developers and ideas for the RealTimeWeb that I've heard yet. I wish FriendFeed has this search engine's capabilities. - Robert Scoble

Okay, you've got me hooked, RS. Gimme more, please, Sir! - James D Kirk

fascinating concept "Fluid Info": "database with a heart of a wiki" ... love his accent! Interesting to give user so much control w/o schemas, permissions, etc. Removes all the structure, limitations....wow - Susan Beebe

ok where's part 2 at? - Susan Beebe

Susan: Uploads from Spain are slow. - Robert Scoble

Faster than China or slower? - David Feng from IM

Sorry - wrong comment... I mean, commented to wrong person... - David Feng from IM

Gotcha! Time for some coffee although that hot chocolate does sound better :) - Susan Beebe

David: I think uploads in China were faster. But I think the hotel wifi sucks. Can't blame that on an entire country. :-) - Robert Scoble

Thanks Robert! And thanks for watching.... I'm a bit of a gasbag. - Terry Jones

Terry your ideas are amazing... ripping away the layers of db constraints to open up an evolutionary platform for information mining / sharing is simply AWESOME! My db structured brain is having a tough time computing the lack of structures, but my intuition tells me this is right for the future....neat data model design concept for sure! - Susan Beebe

Hi Susan. Thanks :-) BTW, there are permissions. You can have, e.g., a susan/rating attribute that you're putting onto things, and it's yours - no one else can detect/read/write/delete it unless you let them. OTOH, the underlying database objects have no owner. There are no permissions at that level. So you can put a rating (or anything else) onto any object (that you can find). No-one can stop you. So we're fully writable like a wiki, but with a permissions structure within the object (unlike a wiki). - Terry Jones

Ooooh got it!!! Ok so you're building a core data model (protected) with objects everywhere (open design) which are available to users who are presented with a highly customizable user profile (detect/read/write/delete ) that can call / manipulate said objects to create the user's *own* data set (which has multiple layers of permissions / attributes / tags? to control sharing, reporting and distribution of data) = fabulous! when can I have it? :) - Susan Beebe

Terry - is this something like entity/attribute/value? I did something like that a few years ago... but found it hard to do searching - equality in my thing was fine, but greater than, less than etc (ie: with a date range) were a bit harder. How have you overcome this? - Brad

@Susan. Yes, that sounds more or less right. Yes, your own data is on the objects, as it that of anyone else who wants to put something there. it's all combinable, searchable as you like. Plus you can organize multiply, simultaneously, and arbitrarily (simply by adding more tags to objs & searching). We'll do an alpha release in early 2009. - Terry Jones

@Brad. Dates are stored both numerically and textually. What gets searched on depends on the query. terry/seen > "Jan 22, 2007" is numeric, terry/seen ~ "Monday" is textual, etc. The query language is dead simple. It took me a lot of thinking to reduce everything to very very simple primitive operations and an easily parallelizable query language. - Terry Jones

BTW, there are 3 more parts to the video coming up... :-) - Terry Jones

Part II is now up: http://www.kyte.tv/scoblei... - Robert Scoble

Impressive demo and ideas. I was thinking about similar lines last week using CouchDB (a schemaless document database by Damian Katz). Also the views in FluidDB sound like views in CouchDB. I very much like the idea of sharing data and being able to annotate or enhance the original(!) data. This is Open Data on steroids. - Berry Groenendijk

Mindnumbing stuff. - Thomas Bøhm

@Berry Hi. I'm reasonably familiar with CouchDB. It's a very different animal. CouchDB is very focused on documents, and lays out complete documents (serialized JSON strings actually, plus BLOBs) efficiently on disk. FluidDB is not focused on anything :-) And its storage is not done by object, but is instead by attribute (or tag if you like). CouchDB used to not have permissions, but I think that's changing/ed. About views yes, agreed. I hope that makes it clearer. It takes a while to get. - Terry Jones

Part #3 is now up http://bit.ly/UteO - Terry Jones

Terry I think I am slowly getting my head around FluidDB. There is still one big problem. The way people tag things. Some people tag things with a x,y coordinates, others with a longitude and latitude, etc. You need consistent tag names (or metadata field names or whatever) to be able to effectively search data. Does FluidDB help you with this in any way? - Berry Groenendijk

And part #4 http://www.kyte.tv/ch... - Terry Jones

Hi Berry. No, there's no help, and nor do I think there should be. Conventions evolve. They become consistent to the degree that it's important they are consistent. If it doesn't matter that you write color and I write colour, then it's no big deal. But if I write S.O.S. and you write S.O.B., it could be very important! - Terry Jones

There's a lot of evolutionary biology thinking behind FluidDB. Attributes will (implicitly) have fitness. Things that are useful will flourish, become trusted, be heavily used, and their owners will similarly gain. Other stuff will not. This gets at the question of spam too. What's spam? But that's another subject - also very important if you're going to build an information architecture that can survive its own success. - Terry Jones

Terry I like the way you think about these things. Just viewed part #4 of the interview. I am looking forward to the alpha release. - Berry Groenendijk

Eager to see Alpha product too! send me invite susan.beebe {at} gmail dot com - thanks! This is really inspiring / disruptive technology!! love it! makes my brain hurt in a GOOD way (i.e. un-doing all the overly burdensome architecture that was imprisoning my data!) :-) Terry is one smart cookie!! - Susan Beebe

How will you expire certain attributes? Similar to domain name management today, will you have leases on groups (or specific) attributes? This seems to be the way that you're going with the revenue model, and would certainly make sense after you've reached enough critical mass. When will you start allowing people to start reserving namespaces/attributes? - Davison

very neat project - mikehedge

@Terry - thanks. I built my thing on a standard relational db - Firebird, but everything was stored as text, so it was a bit difficult... :) So... how do you go about implementing "relationships"? ie: I have a video store, with all these videos, and this person rents these vids, so I'd like to send them an email when there's a new release in the genre that they've hired previously? Would love to talk to you more about this. Thanks! - Brad

@SusanBeebe - you know, that's exactly right! undoing the architecture! yes, sort of kinda..... :) and freeing one'self from the confines of relational theory... but see me other comment to Terry re relationships. Happy Days! :) - Brad

@Terry - sorry I can't stop thinking about this. In my thing, I maintained another table that aggregated all the various uses of a "thing". So at a glance, I could tell what the most used credit card was for purchases.... which state had the most sales etc.... this was a way to help the business owner make sense of all the data, and plan for the future. ie: Diners Club had like 2% of sales, yet attracted the highest fees - so that tells me get rid of Diners Club as a payment method? That sort of thing..... - Brad

@Terry - and all - I had better stop here - I could keep typing all night about this, and then miss the new year! :) So happy new year to everyone. And special thanks to Robert for bringing you to our attention. - Brad

Hi @Brad. Sorry for my slow replies - I don't get any notification of new comments here. I don't really use friendfeed (yet). I'm not sure exactly what @Susan had in mind with her "undoing the arch" comment, but that certainly captures the flavor. Re relationships, there is NO support for them. It's not a relational model. There's just a (conceptually) very simple architecture and laughably simple query language. You get to do analysis on your own CPU :-) Lots of tradeoffs there, of course. - Terry Jones