To me this sounds like an implementation detail rather than something that needs to be supported on the graph level.
If for optimization purposes a server needs to know the data type of a literal (binary, number, boolean, etc.), then it could either
1. figure out the data type based on the dictionary, or
2. require (as a policy) that clients add the literal together with the metadata that specifies the data type
Markus
On Fri, Mar 2, 2012 at 3:05 PM, Michael Schwartz <> wrote:
I think we need a way to differentiate binary literals if we want to support storage of binaries in the graph. And simply storing metadata is not sufficient because there is no guarantee that the metadata is created before the literal needs to be stored. Is there a solution?
thx,
Mike
---------------------------------------------------
Michael Schwartz
Gluu
Founder / CEO
+1 646-810-8761
On Thu, 1 Mar 2012, Drummond Reed wrote:
Mike, the question you raised actually highlights a number of the key
reasons for the literal context proposal. Let me explain.
First, per Markus' last point in his message, one primary purpose of
literal contexts is so an XDI processor always knows the type of the
literal node it is dealing with by the type of the literal context. To use
Mike's example, if "*logo" is defined as a literal context containing a
binary value (which has several semantic issues -- see below), then
@!1111*logo always identifies a single binary value.
Secondly, it's a misunderstanding of literal contexts that there could be
two literal arcs. The whole idea of a literal context is that it has a
single literal arc. If the literal context is defined as containing binary
data, then that literal arc points to binary data. If the literal context
is defined as containing a number, then the literal arc points to a number.
And so on.
These are all controlled by XDI dictionary definitions, which is why I
think the XDI Dictionary spec is moving up in priority.
Now, the semantic issues. First, @!1111*logo is semantically wrong because
uses a *name to identify a literal context defined in a dictionary, and *
is not the dictionary namespace. The dictionary namespace is + (with the
exceptions of a handful of special $words defined by the XDI TC).
So it should be @!1111+logo. The second semantic issue is that +logo should
not be a literal context because it is not a single defined datatype. +logo
should be a supertype that has defined subtypes. In other words, the
dictionary defintion should say +logo/$is$a/+image, and (among other
options) +$image/$a/$a$xsd$jpeg. So if the binary you wanted was a JPEG,
then the address would be:
1. @!1111+logo$a$xsd$jpeg <== canonical JPEG version of the logo
(zero-or-one)
2. @!1111+logo/$a$xsd$jpeg <== (note the slash) relational arc pointing
to all the JPEF versions of the logo (zero-or-more)
So it is the dictionary-defined $jpeg, and not +logo (and definitely not
*logo) that is the XRI literal context storing the binary data. And because
the $dictionary will define $jpeg as a JPEG binary, any time an XDI
processor sees $jpeg as the literal context, it knows the value is a JPEG.
I'll put this on the agenda for tomorrow's call (even though Mike won't be
there).
=Drummond
On Thu, Mar 1, 2012 at 4:35 PM, Michael Schwartz <> wrote:
Markus,
Thanks for the quick response!
1. Yes, you could have both a ! and !! arc, why not? It doesn't seem
inherently problematic to me.
2. yes, you're right I forgot the / for the literal arc :)
3. !! was just an idea... it could be something else.
4. Its useful to be able to differentiate binary data from other types of
data because there persistence strategy varies greatly.
- Mike
1. Could you have BOTH a binary and a non-binary literal in a context?
I.e.
have a ! arc and a !! arc.
If not, will the server throw an error if you already have one literal and
try to add the other one?
2. In your example, wouldn't the XRI actually be @!1111*logo/!! instead of
@!1111*logo!! ?
3. Something is bothering me about the fact that !! is two subsegments. I
am worried it will result in ambiguity in one way or the other.
4. I don't understand why you can't load the metadata together with the
literal, to figure out whether it is binary or not.
Wasn't this the whole point of what Drummond called "literal contexts"?
Markus
On Thu, Mar 1, 2012 at 9:49 PM, Michael Schwartz <> wrote:
XDI TC:
I'm attaching a proposal to represent binary literal values as "!!"
So, for example, the XRI for a my company's logo might be :
@!1111*logo!!
The fact that the image is a binary can't be represented in the metadata,
because it is not known which will be loaded first, the literal node, or
the metadata about the literal node.
This is a critical issue to optimize persistence of binary objects, which
need to be indexed differently.
I'd like to fast track this because we need to store binaries as part of
the oxPlus demo :)
Thoughts?
thx,
- Mike
------------------------------**------------------------------**
---------
To unsubscribe, e-mail: -**open.org<>
For additional commands, e-mail:
------------------------------**------------------------------**---------
To unsubscribe, e-mail: -**open.org<>
For additional commands, e-mail:
---------------------------------------------------------------------
To unsubscribe, e-mail:
For additional commands, e-mail:
