... personal wiki, blog and notes
WCS is dead, long live WFS
For many years Steve Hankin has been asking me why we want WCS when OPeNDAP has similar functionality, and many, many, working implementations. For just as many years I've argued that OPeNDAP has/had three major flaws:
It wasn't easily securable (soon not to be a problem),
Didn't have good relationship with metadata, and in particular,
Was all about syntax, not semantics - you subset by array index, not the desired portion of a semantically described domain (e.g give me array elements 4-6 compared with give me the array elements which lie between latitudes 40 to 60 degrees).
but I've also admitted that WCS had some flaws too:
It might be easier to secure, but only because it (might) be easier to implement your own stack ...
Nearly no working implementations.
You would have thought the latter would be a show stopper, and indeed it is, but the first flaw for OPenDAP has also been a show stopper ... until now. So, we are going to deploy OPenDAP soon ... but we still want to deploy something which addresses semantic subsetting as well. So we've been investing in WCS ...
... but today I heard a presentation that filled me with horror. Very well presented, but still horrific. The plan for the future evolution of WCS is so flawed that I can't see it surviving!
Fortunately, the talk on WCS was followed by one on WFS (declaration: from someone in my group) which outlined how WFS can deliver much of the same functionality as WCS. It remains to be seen whether it can deliver a semantic version of OPenDAP ... I hope so, in which case it'll be "WCS is dead, long live WFS".
So predictions, should either of them read this:
Rob Atkinson will smile: He's been claiming for a while that WCS was, or would be, only a convenience API to a WFS!
Steve Hankin will role his eyes, and think: "oh no, not another WXS ..."