[postgis-devel] Validity flag

[Hugo Mercier]

On 29/02/2016 18:18, Sandro Santilli wrote:
> On Mon, Feb 29, 2016 at 05:57:06PM +0100, Hugo Mercier wrote:
> 
>> What about putting an extra byte at the *end* of the data ?
> 
> Neat!
> 
>> I know it is hacky, but ... can't find better yet.
> 
> The bit (currently proposed for "valid") could actually mean:
> there are more metadata at the end. Why not. I actually like
> it as a way to ensure backward compatibility.

Exactly.

I would still be in favor of being clear about what bits are used/free
and what do they actually mean :) ReadOnly and Version are not clear for
me. As well as the 3 spare bits of the SRID field.

> 
>> It could invalidate memory cache for huge geometries, but should not
>> disturb existing code
> 
> What do you mean here ? Which memory cache ?

I was referring to the L1/2/... CPU cache. Since these extra flags may
be far away from the header. But, humm, that is probably not a big issue.