XML

[John Halleck]

On Thu, 4 May 2000, Michael Lake wrote:

> [...]

> Whats the diff between a covariance for a point and its standard
> deviation for a given instrument?

  What's the similarity?

> Is the covarience a statistical measure of how much it was adjustd by
> the closure?

  The covariance of the final adjusted point is the expected (3d) uncertainty
  of the point.  This is affected by the adjustement, but also by the
  covariance's of the measurements going into the adjustment.
 
> If its not exacly the same then at least it can be treated similarly,
> though perhaps with a new tag if necessary.

  It is *NOT* exactly the same.

> Have a look at the Document Type Definition for the CaveSurveyXML.
> Example DTD "CaveSurvey.dtd" at 
> http://www.science.uts.edu.au/~mikel/cavescript/docs.html
> Below is just a section of the DTD but as you will see whats here 
> [...]

> This means a Survey Series may contain the Surveyors and Instruments
> tags but must contain at LEAST one but possibly more STN and SHOT tags.
> There might be none or more cross sections ie XSECTs.
> The STN tags must have a NAME attribute for each one and may contain
> position information. The dE is error in Easting and likewise for dN and

  The assumption that three numbers give the error, without any measure
  of corrilation is common in the caving community, but not very
  statisticly meaningfull.
  
> dH is error in Height (In fact thats why I don't have a FIX tag - when
> Survex data is turned into this XML format *fixes become STN elements
> with dE, dN, and dH set to zero) 
> The survey SHOT tag in any CaveSurveyXML document must
> contain the atttributes DIST, BEAR and ELEV (ie D, A, I) and optionally
> standard deviations from these values ie delta DIST which I call dDIST
> etc. 

 The only way this makes any sense to me is if you are assuming that
 there is never any corrilation between the measured values.

> > I wouldn't mind having (but it would probably clutter the world) the
> > ability to include (so that updates are minor) generated analysis
> > data such as reference counts, and the spanning tree of the
> > geometry matrix.
> 
> Uh? I've a lot to learn here. 

   A spanning tree of a survey is a maximum collection of shots that
   ties all the shots together without any redundancy.  If you have
   a spanning tree, then each remaining shot of the survey is associated
   with one (and only one) elementary cycle of the graph.

   Spanning trees, and minimal spanning trees, are quite usefull in
   analyzing a survey...  And once computed are easy to add to.

   Almost any book on Graph Theory will discuss spanning trees.

> >     ---------- Forwarded message Two ----------
> > On Wed, 3 May 2000, John Halleck wrote:
> > 
> > Almost any program that does any analysis of the data is going to
> > have to assign a unique number to each survey point.  If this could
> > be put in the data, then further analysis programs could use the
> > previous numbers.  (The same could be said of shot numbers...)
> 
> Not needed in most cases. In XML one can have unique ID's attached to
> tags but how I want to refer to other survey shots or stations is with
> XML's Xpointer and Xlink constructs. Thats how I want to convert
> imported Survex data *equates when I use my svx2xml script.
> Just refer to all stations by their name. In cavern its say

  In order to do an adjustment of the network, one forms a matrix.
  Matrices are traditionally subscripted by numbers, not by names.

  I've never seen an adjustment program that subscripted (for example)
  a shot list by anything other than integers.

> /mycave.pitch1.extension.5 whereas in my XML format I nest the survey
> series. My DTD allows this as so:
> <!ELEMENT SERIES ( (SURVEYORS?, INSTRUMENTS?)?, (STN+, SHOT+)+, XSECT*,
> SERIES* >  
> A SERIES can contain ie zero or more other SERIES. An example is
> example2.xml on my page.  
> 
> > If there is a range used put somewhere up front, then combining
> > sections could be made efficient.
> 
> Again use the powerful Xlink and Xpointer so then linking surveys is
> like linking HTML documents. But unlike HTML you dont need to insert
> anything at the place where you want to link to but thats another story. 
> 
> Mike
> 
> --------------------------------------------------------------------
> Michael Lake
> University of Technology, Sydney
> Email: mailto:Mike.Lake@uts.edu.au Ph: 02 9514 1724 Fx: 02 9514 1628 
> URL: http://www.science.uts.edu.au/~mikel
> Linux enthusiast, active caver and interested in anything technical.
> --------------------------------------------------------------------
>