This sounds a lot like ASN1 encoding - maybe we can get
some work for free here. Maybe a relaxed form of ASN.1 - BACnet is like
this, it's not a full ASN.1 but it uses the parts of ASN.1 that are useful for
it and skips it in other places.
From: Brian Frank
[mailto:]
Sent: Wednesday, August 19, 2009
3:53 PM
To: Considine, Toby (Campus Services IT)
Cc:
Subject: Re: [obix-xml] Re: Proposal for
Alternate oBIX Encodings
> or are you looking at
something like an ASN1 encoding…
My suspicion is that a
oBIX specific encoding will give us much better compression and be far easier to
implement. For example we have a budget of a few KB in Sedona. Right
now I am thinking about
- a byte code per
element type with implied value encoding
- a bit in the element code to
determine open/close element (children elements)
- a byte code per attribute with
predefined value encodings
Because the
element and attribute names are fixed, byte codes can all be predefined for
everything.
But we need to implement it and
see how it compares to other strategies such as Fast_Infoset.
The good news is that no matter
what we settle on, I think we can get 3 different open source implementations
done quickly:
- Java oBIX toolkit
(http://sourceforge.net/projects/obix/)
- Sedona (http://sedonadev.org/)
- Fan
(http://fandev.org/doc/obix/index.html)
Brian
