Re: Hungarian data for read-write apps

Hey Michiel,

Interesting, so this would basically be an alternative to the RDF data
model or other semantic data models, right?

I think it's great, but here are some ideas:

* It took me a while to figure out what you meant by "hash". First, I
thought it would mean that the value of that attribute is a cryptographic
hash of something. Only then I realized that you meant hash in the sense of
e.g. Perl, i.e. an associative array. Maybe a better word would be "map" ?

* In many data models I am familiar with, there is a distinction between
"relations" (to other things) and "literals" (strings, scalars, booleans,
etc). But I kind of like how you say everything is a "relation" between
"things".

* I think you use the word "statement" ambiguously. First, you say we can
describe the real world as a set of "statements", each one of which
contains a "relation" between "things". But then later you use the word
"statement" to refer to boolean objects.

* Instead of "document", why not just say "string"?

* I don't think you ever mention the notation for a document/string that is
not an encoding of an object. E.g. how would you express that my first name
is "Markus"? Would that be what you call human-readable language ("h") ?

* Your use of the words "thing", "object" and "variable" was also a bit
confusing to me, but only when I read it first, at the second time it all
made sense :)

Anyway, I think it's a great approach.
I guess terminology is always a big challenge when defining data models.
The next step would probably be to define operations on the data!

Markus
-- 
Project Danube: http://projectdanube.org
Personal Data Ecosystem Consortium: http://personaldataecosystem.org/

On Sat, May 12, 2012 at 11:40 AM, Michiel de Jong <michiel@unhosted.org>wrote:

> read-write web turns the web from a world wide "book" into a world
> wide "software program". this means we can now interact with it not
> only to consume information, but also to push buttons that record
> (editing the web), as well as effectuate (APIs on the internet of
> things), changes in the real world things described by the web's
> documents.
>
> i've been thinking about how to model this in a way that reduces the
> number of potential bugs ('word jokes'). if you're subscribed to the
> public-tag cg, you may have seen my thread there about distinguishing
> various types of URLs in our software. this train of thought has lead
> to this sunny Saturday morning coming up with an Apps Hungarian
> notation for local variables in javascript applications that
> manipulate the read-write web, as well as for attribute names in
> machine-readable documents on the web. Even if you don't jump to adopt
> it, it might be an interesting straw man argument about how we should
> name things in linked data in general, and on the read-write web
> especially.
>
> The only thing i love more than computer science is metaphysics, so
> let's start with some of that:
> The way we will model the real world today is as a collection of
> objects (our variables; not necessarily physical objects, but any
> 'thing') that have certain relations to each other (the relations are
> what makes for our actual data). We can describe the real world with
> an array of statements, each of which says that thing X has relation R
> to thing Y. Since objects are only variables, to represent them we
> need to talk about how they relate to other objects. That's why in
> computer programs, objects are represented as lists of relations to
> other objects. so if X is the 'parent' of Y (without specifying
> further what that means), then we can do:
>
> var Y;
> var X = {
>  parentOf: Y
> };
>
> That's the basis of our metaphysics here. Now we can make it a bit
> richer, by saying that an object can be zero or one of the following:
> document (string), statement (boolean), scalar (number), array.
>
> If an object is a document, then it has contents (which may be the
> empty string). If a thing is a statement, then it has a truth value
> (which may be undecided or unknown in the current world view)). If a
> thing is a scalar then it has a numeric value. If a thing is an array,
> then it has zero or more element objects in it. Simple, right?
>
> Let's add one more useful data structure that computer scientists like
> to use when describing the world: the tree. A tree has a root, nodes,
> and leaves. The root points to nodes or leaves, nodes point to nodes
> or leaves, and the leaves are objects. Actually an array is a special
> kind of tree, but let's treat them separately because of how arrays
> are used in software.
>
> To talk more about representation, we already saw that an object can
> be represented by a variable, but it can also be represented by a
> document, using an encoding. Example encodings are JSON, XML, and
> human-readable language. Normally a document is either read-only or
> read-writable for the current user. This affects what we can do with
> this document in a 'read-write app' context. We also allow documents
> to be append-only, like printer queues. In that case we call them
> virtual documents.
>
> Now we can introduce our Hungarian notation to deal with these
> documents, statements, scalars and arrays, as well as with plain
> objects.
>
> attributes hash of the object:
> _    If a variable holds a hash of attributes of object 'thing', then
> we give this variable the name 'thing' (camel case).
>
> document describing another object:
> j    If a variable holds the contents of a document that describes
> object 'Thing' using JSON, then we prepend a 'j', and use camel-case,
> so: jThing
> x    same for xml
> h    same for human-readable language
>
> virtual documents forming a way to contact another object:
> c    for instance, a mailbox or a printer queue on the internet of
> things is an appendable document (this is a bit abstract, but it helps
> simplify our model of the web and the APIs that live there if we think
> of message queues as just documents). if you have a variable in a
> program where you accumulate messages for Person, but that does not,
> as a variable, represent that person, then you would call this
> variable cPerson.
>
> boolean describing the truth value of a statement:
> b    if a variable holds a boolean describing whether 'statement' is
> true in the current world view, then we call this variable bStatement
>
> number describing the value of a scalar:
> n    if a variable holds a numeric value representing the value of
> 'scalar' in the current world view, then we call this variable nScalar
>
> data structure containing objects:
> a    if a variable holds an array of objects, then we can name that
> variable aObject. for example a variable holding an array of 'author'
> objects would be called aAuthor.
> t    if a variable holds a tree (arrays, nested recursively), then we
> can name it after its leaves, so if the leaves are authors, then we
> call the tree variable tAuthor.
>
> web presence:
> p    if a variable holds the URL of document 'document' then it would
> be confusing to call that variable 'document' (especially because both
> documents and URLs are strings). so then it's better to call that
> variable pDocument.
>
> So that's the Hungarian notation that i'm proposing for the read-write web:
> - no prefix for attribut hashes,
> - j,x,h for strings describing objects
> - c for 'virtual documents' that are contact interfaces of objects
> - b for truth value of a statement
> - n for numeric value of a scalar combinations
> - a,t for composite objects
> - p for strings (addresses) describing where to obtain/edit documents
> or where to interact with virtual documents
>
> So we could make combinations obviously, to get more powerful naming
> schemes.
>
> For instance, applying Hungarian notation to both variable names and
> attribute names, how would you Hungarianize this:
>
> var book = {
>  authorMailboxUrl = 'mailto:user@host'
> };
>
> The variable name 'book' is OK because the variable is an attribute
> hash of  a book. But obviously, the string 'mailto:user@host' is not
> an attribute hash of the author. It's an address at which we can
> interact with the mailbox of the author. That's why this attribute
> name is so long -  'authorMailboxUrl' - a really cumbersome attribute
> name. So what a lot of (IMO sloppy) programmers would do is just call
> the attribute 'author' and then say 'well, programmers that come after
> me will know what i mean'. But luckily, we no longer have to be so
> sloppy, and can now abbreviate it, first to 'pAuthorMailbox' and then
> even to 'pcAuthor'.
>
> That's my proposal for using Hungarian notation in linked data. i
> think it's absolutely necessary to introduce this. i'm going to start
> using it myself. It will probably change as i start using it, this is
> just a first draft. Let me know what you think!
>
>
> Cheers,
> Michiel
>
>