[Matroska-devel] float vs. double issues

[Matthew Heaney]

> -----Original Message-----
> From: matroska-devel-bounces at lists.matroska.org [mailto:matroska-devel-
> bounces at lists.matroska.org] On Behalf Of Moritz Bunkus
> Sent: Tuesday, 30 March, 2010 15:21
> To: Discussion about the current and future development of Matroska
> Subject: Re: [Matroska-devel] float vs. double issues
> 
> And you're right: EBML floats can be
> either 'float' or 'double' C data types. The only floating point data
> type that must not be used is 'long double' (the 80bit type).

> When writing the application has the choice between 32 and 64bit
> floating point types. The reader has to support both.

How is a parser supposed to know what the serialized type is?  By inspecting
the size field?