Paul Sandoz wrote:
> Alan Hudson wrote:
>
>> I've gotten an early prototype of the Xj3D / FastInfoset integration
>> going.
>> Using a 20 meg CAD file
>>
>> Size(bytes) Parsing Time(ms) Speed
>> XML 20396657 10406 1 :1
>> FI, Strings 20340458 5672 1:83 :1
>> FI, builtins 11889015 1546 6.73
>> :1 !!!
>>
>> Only near 700% speed improvement on parsing. Nice work!
>>
>
> Great.
>
> Are you also binding the string data after parsing?
>
yes, data is bound after the string parsing. All time is included in
this chart. Ie the time it take to load the file, and parse into the
applications space.
> I am wondering why 'FI, Strings' increase is not greater then 'XML'
> and only smaller by 50K. If the 20 meg file contains mostly very large
> attribute values and element content then i can see why this would be
> the case.
>
the data is almost all large attribute data in this file.
> A little tool caculating the percentage of markup to content would be
> very useful.
>
>
>> I'm going to start adding our X3D specific encoders now. I expect
>> they will help filesize but likely harm parsing speed.
>
>
> We should also compare with GZIP XML and GZIP FI.
>
agreed. when I put FI into the XSBC file comparison panel it will give
us these comparisons. Right now all files are becoming smaller GZIP FI
then GZIP XML.
>
>> I'm also going to prototype a BYTE encoder for possible native
>> inclusion into FI.
>
>
> OK. We need to standardize the URI for this algorithm. Will this be
> included in the X3D algorithms?
>
>
I feel its more general then that. But I'm also happy to have X3D get a
little more publicity by having this in its namespace. Your call.
>> I expect I can release a public version of this in a few days. Feel
>> free to share these results, I expect to make more results available
>> once I finish the implementation.
>
>
> Great. We will be presenting a paper to XTech 2005 at the end of May
> and some X3D numbers would be good to show off the use of encoding
> algorithms (built-in and application).
>
I will be showing this off at the Web3D conference March 28. And if we
get into Java One then I and/or Don will show it there.