[postgis-devel] Canonical binary form AND WKB

[strk at refractions.net]

Markus, I'm not convinced this is the right path to take.

If we made the canonical form == internal form we'll have
many troubles when willing to change internal form and 
accept restores from 'old' dumps.

Also, needs of internal storage are different from needs
of data exchange. For internal storage we want a representation
which is performant and small in size. For exchange we want
a representation which is stable, essencial and portable.

For example we do not need BBOX when dumping.
Similarly we do not need endiannes in internal storage.

Another example is that the first issue (dumps compatibility)
could be solved adding a version number to the representation,
but we really do not need it in internal storage.

--strk;

On Mon, Dec 20, 2004 at 03:35:05PM +0100, Markus Schaber wrote:
> Hi, strk,
> 
> On Sat, 18 Dec 2004 10:41:41 +0100
> strk at refractions.net wrote:
> 
> > That would be the best solution, as it seems to be getting a mess
> > with all these representation:
> > WKB, EWKB, HEXEWKB, internal, canonical binary, WKT, EWKT
> 
> Then lets try to get rid of some of them, and simplify the others.
> 
> Add the endian bit to internal representation, so internal and canonical
> binary get the same. (Always convert to machine byte order on geometry
> input so we do not have any speed issues, and do no conversion on
> output. Maybe we should make sure that we have space reserved for some
> additional bits for future extensions).
> 
> Make canonical text a straight hex(canonical binary), instead of
> adding the SRID with a ';'.
> 
> So I see four "real" representations left:
> 
> - internal == canonical binary == unhex(canonical text).
> 
> - OGC WKB 
> 
> - OGC WKT
> 
> - Old SRID=42;EWKT format, for canonical text input only
> 
> Canonical binary should be as efficient as possible. We cannot abandon
> the OGC formats, obviously, and the old format has to be maintained for
> restoring old dumps.
> 
> Old format can use the same parser as WKT after splitting the SRID part.
> 
> Maybe the canonical binary and WKB parsers can share some code,
> especially endianness conversion.
> 
> This approach should be the leanest.
> 
> But I'm pretty shure I forgot some important use case :-)
> 
> HTH,
> Markus
> 
> -- 
> markus schaber | dipl. informatiker
> logi-track ag | rennweg 14-16 | ch 8001 zürich
> phone +41-43-888 62 52 | fax +41-43-888 62 53
> mailto:schabios at logi-track.com | www.logi-track.com
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-devel