Binary RDF

[Robin Berjon]

Hi all,

On Aug 18, 2011, at 20:45 , Kjetil Kjernsmo wrote:
> On Thursday 18. August 2011 18:47:42 Toby Inkster wrote:
>> I've come up with ideas for such formats, but on examining each idea
>> have come to the conclusion that the search for an ideal binary RDF
>> format is still incomplete.
> 
> Yeah, and with competing needs, it isn't given that such a thing exists. 
> Anyway, I would certainly like to hear Robin's comment, are you around?

Yes, I am, just back from vacation :)

I am not sufficiently familiar with all the use cases for binary RDF to make an informed judgement about whether the submission addresses the broader needs of the community or not, so I'll simply make some general comments based on the long, long binary XML odyssey. I'll keep them short, but don't hesitate to ping me with questions — I would hope that one binary serialisation project could learn from the experience of another!

The first thing I'll note is that if you want to make a truly generic serialisation, you probably can't skip the process that the XBC WG (XML Binary Characterisation) went through. First, we wrote three documents in parallel: a Use Cases document (http://www.w3.org/TR/xbc-use-cases/) in which every use case had a list of must/should/nice properties it required from a format; a Properties document (http://www.w3.org/TR/xbc-properties/), which defines said properties; and a Measurements document (http://www.w3.org/TR/xbc-measurement/) which explained how to measure each property so that different formats could be concretely tested. Based on these, we then had all the material needed to ask all the stakeholders in the community which properties, were they to be *absent* from a given format, would cause them to absolutely not be able to use it and force them into using something else (this is phrased negatively to avoid the shopping list effect). This allowed us to write the Characterisation document (http://www.w3.org/TR/xbc-characterization/) which had minimal requirements on a format. The format was then elaborated: http://www.w3.org/TR/exi/.

I know that this sounds terribly convoluted (and it is) but there's hope nevertheless. First, you might be happy to know that it only took a year (to the day) for this to be finished, which in standards time is lightning fast. Second, RDF is different from XML and it tends to be bent backwards less often (in part because it's designed to bend). As a result I would expect fewer use cases (in some cases, I also expect that you can reuse the ones that are there). Third, you can probably reuse many of the properties. Fourth, it's likely that you can skip on much of the measurements work — that is mostly useful if you have many proposals in a hotly contested space, and if you are required to prove that your tech can work to a large group of people who don't believe it's useful at all in the first place. So all in all, it's probably not *that* much work.

The second thing I wanted to note, and which might also bring some hope, is that it's not impossible that the EXI format could actually be reused for RDF. The X in there does indeed stand for XML, but it is in fact a very versatile format that does not require XML to be involved. The underlying model is the XQuery data model, which means it can capture many things that XML cannot — notably sequences or multi-root documents. It wouldn't be surprised if a decent mapping could be used for RDF that might correspond to a subset of EXI. It's certainly worth looking into it.

-- 
Robin Berjon - http://berjon.com/ - @robinberjon