[Prev][Next][Index][Thread]
Re: re Optimisation / Sparse Matrices (plain text)
> I have an important concern: Are these
> standards going to be appropriate for communicating huge objects
> (matrices, etc...)?
Yes and no.
For archival purposes, and for communicating between systems
that have no other joint language, converting even very large structures
into an XML encoding (OpenMath or MathML) may make sense.
On the other hand for two processes that are sharing some gigabyte
matrix computation, clearly they are going to pass binary data, the
draft openmath standard explictly advocates three levels of
comunication. (See figure 2.1 in the introduction). Processes may share
internal private representations (if they have the same representation)
or they may share internal versions of the OpenMath object, or they may
communicate via an external representation of the object (XML or binary
encoding for OpenMath). The situation for MathML is broadly
similar.
Also, for manipulating MathML objects without serialising to an XML
linear representation, the MathML2 drafts have introduced the MathML DOM
specification (Chapter 8) that should provide a standard API for
manipulating these objects.
David
(wearing both MathML and OpenMath hats)
References: