[Xml-bin] Status of projects?

[Al Snell]

On Tue, 5 Jun 2001, Anders W. Tell wrote:

> What is your take on the slightly different usecase with IPC
> (interprocess communication) where the information transferred
> originates from objects.
> Isnt it more efficient to use pure Corba (CDR) or native ASN (BER,
> PER,..) than XML encode the information first and then do a PER encoding ?

Yes, but that doesn't include The XML Buzzword(tm) That Obviously Ensures
This Product Is Not Monopylous And Will Interoperate Seamlessly And
Automatically With The Semantic Web(tm), if you know what I mean...

Ideally, the XML APIs would be implemented on top of the PER encoder so
there's no actual text-XML phase, but the ability to chuck in an "adapter"
as a quick hack will help development, I guess.

> >The result
> >will still be amenable to gzip if bandwidth is more crucial than CPU power
> >and memory, since it will contain all the CDATA as-is, and CDATA generally
> >compresses well.
> >
> Interesting, it will be great to see your final solutions.
>
> BML differs from ASN based encodings in that NO knowledge of schemas is
> required, the focus is clearly on the  end-to-end performance aspect and
> it is based on CDR.

Yup. There's definitely demand for that too!

> Im have however added a GZIP based  character compression mechanism but
> the resulting compression doesn work as one might expect. What I did was
> to separate characters from markup and encode the GZIP'ed characters
> into the stream, but GZIP'ing the whole BML stream is usually more
> size-effective.

Quite - the element boundaries usually follow repeated patterns too.

And if you *individually* gzip each string, you lose out a lot on the
per-gzip-overhead...

> Cheers
> /anders

ABS

-- 
                               Alaric B. Snell
 http://www.alaric-snell.com/  http://RFC.net/  http://www.warhead.org.uk/
   Any sufficiently advanced technology can be emulated in software