SRS handling questions

[Martin Schulze]

Moin!

Jan-Oliver Wagner wrote:
> > I wonder how I should interpret the WMS specification with regards to
> > SRS handling.
> > 
> > The specification [OGC 01-068r3, §7.1.4.5.5, page 26] says about
> > inheritance of SRS:
> > 
> >  - Every layer shall have at least one SRS element that is either
> >    stated explicitly or inherited from a parent layer.
> > 
> >  - Layers may optionally add to the global SRS list, or to the list
> >    inherited from a parent layer.
> > 
> > Since the specs don't lose a word how these two cases are
> > distinguished, I'd consider them clashing.
> 
> I do understand the first rule and the syntax should be clear.
> To understand the second rule I would like to see an example to
> understand how this rule is to be expressed in the syntax.

*smile*  If I had an example and a different syntax, I wouldn't have
to ask.

> > Does anybody feel I should not implement the first rule and skip the
> > latter?
> 
> Are you talking about the 2 rules or the 2 cases of the second rule?

Since I don't know when which of these rules may apply since I don't
know how to distinguish between <SRS>-replace and <SRS>-addtoinherit
I'm talking about the two rules.

I guess the two cases of the second rules could be aggregated to
'add SRS to inherited list'.  However, it's not very clear in the
document as well.

> > Also, since the XML response has both <SRS>foo</SRS> and
> > <BoundingBox SRS="foo" minx=...> elements, I wonder if it wouldn't be
> > wiser to extract information about availibl SRS from the <BoundingBox>
> > element, except for the root layer.
> > I'd say yes, but I'd better ask if somebody sees a problem with this.
> 
> Are you talking about XML responses that came from the frida demo wms?
> If yes, please note that it is an older UMN MapServer and not
> necessarily 100% compliant with WMS specification.

I already noticed.  According to the specs the root layer SRS is wrong.
However, applying the programming paradigma below it doesn't matter
to our implementation.

However, that wasn't the reason, even though I'm currently using frida
as an example.  I'm merely talking about a good programming paradigma
versus a strong specification implementation: "be strict in your output
but lax in what you accept".

> If there are both, they should be the same, right?

Yes.  They should.  However, if they aren't, and the list of SRS is
larger, it's likely to cause problems since there won't be a bounding
box associated to an SRS.  Hence, I tend to ignore the <SRS> tags
except for the root layer, and generate the list of SRS for child
layers from their bounding box elements.  Except if you disagree.

> So, at least there should be a test whether this is the case and
> raise a warning if not.

This could be achieved.

Regards,

	Joey
-- 
Computers are not intelligent.  They only think they are.