Hyperfiddle's Data Gravity, Martin emails
The user pitch is "this tool lets you make sophisticated database apps by yourself", that's the whole pitch, the semantic angle doesn't matter.
The investor pitch has to explain growth, which is unfortunately sophisticated:
This tool also lets you mashup data from existing apps into new apps. Making apps and mashing apps is the same tool and the same process. All our documentation will be in the form of live mashups so they will figure it out. You have an idea that needs to use data from some other app, you mashup the data into a new app, the mashup collects its own additional data. So the data attracts more data like gravity. It turns out that this is data network is the semantic web but it doesn't even matter. I should stop talking about it.
Our growth depends on how Easy the mashups are and how strong the Gravity is. For example Facebook's api has huge gravity that attracts thousands of other apps. Facebook's gravity scales with the number of people in the social graph. Hyperfiddle's gravity scales with the amount of total data across all apps connected to it. Hyperfiddle is a black hole sucking in all these apps and increasing its datamass with each one. When compounding interest even a small initial principle, given enough time has vast returns, and Hyperfiddle's interest rate scales with data and grows forever, which is much stronger than Facebook's interest rate which scales with number of users a finite number.
My problem is that I have to manually evangelize a small initial principal, and optimize the shit out of ease of use, hit the critical datamass threshold so the thing collapses into a black hole and the hockey stick growth happens. And I don't know how long that's going to take. Will I have to build Excel-for-Datomic first because that will take $2 million? I am staring at this chasm shrouded by fog.
There's only one world wide web, its gravity is so strong that companies and governments participate, it's a new frontier. There will only be one semantic web of data, and if you make an app that depends on data from the web, you must participate, the gravity is 9999999. The semantic web is the final frontier, the foundation of a new era of computing.
- On Sun, May 21, 2017 at 2:22 PM, Martin Snyder wrote: Does the data need to have any particular characteristics? I can convince myself that Waze, JIRA and Zillow are all sophisticated database apps. Is one better suited to Hypercrud?Is the "sharing" notion of any importance to you? Before, I remember this being almost essential, but now I'm less sure. Would I want to use this if you take "data gravity" out of the equation?Does the data being mashed-up get copied into your system or does it remain outside the system? What's the prototypical data source that is an "existing app." Have you already built a hypercrud app that utilizes it? If not, why not?This is all very well articulated. I'm still stuck trying to imagine a concrete example.
Here is xray vision into dustingetz.com, which is a mashup. Blue is dustingetz.com database. Green is a third party user database. This mashup's query joins across them securely. A 2017-era api cannot do efficient database joins across third party APIs.
Green and blue actually are separate databases. Today we have to host them all, but it will work distributed someday for example if you have to host your own database with validated processes or whatever. This probably doesn't make any sense because you are thinking in sql databases (centralized) but in immutable architectures like datomic and git (distributed peers) it makes sense.
So imagine I need to depend on data from the green user database, so i am forced to use hyperfiddle to get at it, and now my blog app has its own data too which is blue, the blue data also has to be in hyperfiddle because the whole point of using hyperfiddle in the first place is to efficiently join against green. Gravity.
Though yes if you take away the sharing, you're left with "just another open source framework", it is very state of the art and i anticipate we will see a lot of early adoption just for this reason. If we can get to that point in the next year, things will be great. We'd essentially be competing with Meteor at that point, which is a perfectly successful company even without any gravity.
- On Sun, May 21, 2017 at 4:45 PM, Martin Snyder wrote: So how do you do an efficient join across a third party API without having a local copy of it? In that metaphor, is only the "blue" data source writable, or can you write to green data sources as well?
The process where the query runs does have a local cache of database segments; it is a distributed peer architecture like git. So that means, queries in git run on the machine asking the query against a cache, this machine is a "peer" and peers scale horizontally.
Green and Blue are third parties of each other. They both of course have the hyperfiddle interface which is all the API you need since security happens at data layer, not api or query. Security is a filter of the data timeseries, the filter predicate is defined by the owner of the database.