Re: [Tracker] Extracting Embedded Licenses

[jamie <jamiemcc blueyonder co uk>]

On Mon, 2007-06-18 at 12:49 -0700, Jason Kivlighn wrote:
> Hi,
>
> imagemagick: Uses 'convert filename xmp:-' to output an image's embedded
> XMP.  This works for at least JPEG and TIFF files.  For JPEGs, however,
> Imagemagick outputs the namespace and XMP, seperated by \0.  I'm not
> sure how I can handle this, without simply assuming that 'convert'
> returned two null-terminated strings.  Nevertheless, this extracts the
> XMP from TIFF files.

Thats a bit crappy of imagemagick - its too dangerous to look past the
\0 as we dont know the length of the entire string thats returned

Could you use libexif for jpegs here?

> msoffice: Extends the msoffice extractor to also parse the
> DocumentSummeryInformation infile, which contains user-defined metadata,
> along with license metadata embedded by the MSOffice Creative Commons Add-in
>
> pdf: Extends the pdf extractor to read a PDF's metadata stream and parse
> it as XMP.  I'm still awaiting poppler extending the glib bindings to
> allow reading the metadata stream.  Until then, it will simply never
> find the metadata stream and go on without error.
>
> png: Adds a check for the XML:com:adobe:xmp iTXt field, and parses it as
> XMP.
>
> html: Adds a new html parser using libxml2.  Parses the document,
> checking for RDFa licenses.  It also checks for other basic HTML
> properties like title and author.
>
> There's also several XML formats I'd like to parse for license data,
> particularly SVG and SMIL.  Would this be do-able, and if so, how should
> I go about it?  Write new extractors for each format or is this too much
> overhead?  These could use GMarkupParse, rather than bringing in libxml2
> like the HTML parser.


these look ok - will apply as soon as the other stuff is tidied up (need
tracker_read_xmp before I can apply these)

is the reading of xmp just limited to license info or can we extract
additional stuff? I believe the xmp format defines expected metadata for
images so it would be nice if we could fully exploit these.

If you are using libxml for html metadata then I suppose its ok to use
it for the others

I have been thinking of splitting tracker extract into serveral
independent executables to prevent slowdown from linking so many libs


jamie