Re: Domain Specific Languages

[Benjohn Barnes <benjohn@...>]

On Tuesday, September 10, 2002, at 09:29  am, Reuben Thomas wrote:

> > <OT>
> > To think MS chose XML as native format for Office 11... Unless they 
> > always
> > compress it, I expect the files to bloat excessively. Hard disk is 
> > cheap these
> > day, but I still have a 6GB HD on my personal computer, and 
> > applications
> > already eat a lot of it...
> > </OT> -- See, even mailing list/newsgroups notations are infected :-)

I'd far rather they did that than the binary brain dump they usually 
choose. Binary files are dreadful for forwards, backwards, and external 
compatibility.

> That's really just a question of how you encode XML. If you store it as
> ASCII, yes it will bloat (but Gnumeric does this, and it compresses 
> well:
> a spreadsheet I have is ~120Kb of bloated XML, but only 2.5Kb when
> gzipped).

Compressed XML is an efficient storage format for general data. The 
repeated tags are exactly what zip like compressors exploit. I'd 
imagine that compressed XML could be loaded and decompressed a lot 
faster than uncompressed XML could be loaded.

I recall hearing about research in to processing directly on entropy 
compressed (zip like) files. With that approach, it may be possible to 
parse and search directly on a compressed file, giving tight storage 
and fast processing.

Cheers,
	Benjohn