SCS spec bug

[Markus Demleitner]

Hi DAL,

With SCS-1.03 Erratum 2, there is now perhaps less pressure to update
the SCS spec, and hence we should perhaps consider fixing the example
error response at about 48% of
<https://ivoa.net/documents/REC/DAL/ConeSearch-20080222.html>.

Currently this says as the preferred option:

  <?xml version="1.0"?>
  <!DOCTYPE VOTABLE SYSTEM "http://us-vo.org/xml/VOTable.dtd">
  <VOTABLE version="1.0">
    <DESCRIPTION>MAST Simple Cone Search Service</DESCRIPTION>
    <INFO ID="Error" name="Error" value="Error in input RA value: as3f"/>
  </VOTABLE>

Even when adding an xmlns for VOTable, that's invalid, as VOTABLE
needs at least one RESOURCE child.

Since I'd expect we'll go along with SCS 1.03 for a bit longer, what
if we issued an erratum that fixed that example to

  <?xml version="1.0"?>
  <VOTABLE xmlns="http://www.ivoa.net/xml/VOTable/v1.3">
    <DESCRIPTION>MAST Simple Cone Search Service</DESCRIPTION>
	    <INFO ID="Error" name="Error" value="Error in input RA value: as3f"/>
    <RESOURCE/>
  </VOTABLE>

(the namespace is *the* VOTable version 1 namespace for now and
eternity).  That has also dropped the DTD declaration; we should
probably use the opportunity to remove this in the other two places
it's still there, even more so since the DTD URI now resolves to
something that would badly confuse any XML machinery still
frivolous enough to pull DTDs from the net.

For clarity, it would probably good if we said right above the
editor's note on the error responses something like:

  Note that a VOTable element needs at least one RESOURCE element.
  Hence, when the error element is a direct child of VOTABLE, a
  (probably empty) RESOURCE element still needs to be present in the
  response.

Of course, we could also go all the way and remove the parenthesis
from:

> If an INFO element is used, it must appear as a direct child of one
> of the following elements:
> 
> * the root VOTABLE element (which is preferred by this document), or
> * the RESOURCE element

and then add something like "which is preferred in order to bring SCS
somewhat closer to modern (DALI) practices" to the "RESOURCE element"
option.  I feel that's a rather complicated change for little gain
-- we won't make SCS errors DALI-compliant this way anyway --, but I
could be swayed.

Opinions?

         -- Markus