[Psidev-ms-dev] problems with using IEEE-754 in mzData

[Geer, L. (NIH/NLM/NCBI) <le...@nc...>]

Hi,

I've only recently read the mzData standard and don't know what state it is
in, so I hope you will excuse the following comments.  

From an implementation viewpoint, the use of the IEEE-754 floating point
representation to specify the spectra creates several issues:

1. Uneven implementation of IEEE-754 on various platforms:  For example,
IEEE-754 allows "denormalized" floating point values.  It is my
understanding that this subset of the standard is implemented on some
platforms/compilers (e.g. windows) and not on others (e.g. mac).  So unless
an extensive conversion API is written, a float point exception or other
problems could arise from files moved across platform.

2. The IEEE-754 encoded spectral data can not be validated by an XML parser
and is not directly accessible from XML parsing api's.  As far as an XML
parser knows, the spectral data is an opaque blob.  If the data were stored
as XML floats or doubles, the XML parser could validate the data and provide
quick access to the individual peaks.  Otherwise the validation and access
is pushed back into a custom API that may not be readily available and may
have to be written from scratch.

3. A human can't read the IEEE-754 blob.  This may seem a trivial issue, but
it *really* helps in debugging, particularly with end users, to be able to
read the file.

Obviously, putting the data in XML float or double format will address these
issues at the expense of file size.  Gzipping XML can be a practical way to
minimize the file size.

Regards,
Lewis