Re: [h5md-user] H5MD: h5md/author attribute

[Peter Colberg]

On Tue, Aug 09, 2011 at 02:16:01PM -0400, Pierre de Buyl wrote:
> Le 26 juil. 11 à 10:24, Peter Colberg a écrit :
> 
> >On Tue, Jul 26, 2011 at 10:14:05AM -0400, Peter Colberg wrote:
> >[…]
> >>I propose the "author" attribute should contain the full name
> >>(not the
> >>login name, which might be considered “confidential” by some users?).
> >
> >Or better, we could specify use of the full name if available, and
> >otherwise (if the gecos field is empty) fall back to the login name.
> >
> >>On the implementation side, POSIX systems provide getuid(2) to query
> >>the numeric user ID, and getpwuid_r(3) to query the real name (in
> >>pw_gecos field).
> >>
> 
> As far as the implementation is concerned, we shouldn't care too
> much in the specification.

I only mentioned the implementation details to show how to
retrieve the full name of the user running a simulation.
This is not intended for the specification.

We could add separate tutorials (for C++, Fortran, …) which provide
hints on how to implement the specification. For example, I assume
new H5MD users will commonly be new to HDF5 as well, so examples
might be very helpful.

> However, proposing a default seems fine.
> I would go for
> Real Name [,Organization][, <address@hidden>][URL]
> where the [] mean optional. An email or homepage allow people to
> designate themselve in a flexible way.
> This would remain a single text field.

The (optional) email is a good addition, however the comma notation is
weird, and reminds me of broken mail clients that set the From: header
to “Last name, First name <…>”.
I prefer the standard convention of “Real Name <address@hidden>”,
which is e.g. used to indicate author and committer of a git commit.

The organization need not be specified in the specification of the
author field. If a user wishes to include it or any other additional
information *as part of* Real Name, that is fine. Real Name is an
arbitrary string (in UTF-8?). In most cases the organization may be
derived directly from the e-mail address, which makes this information
redundant anyway.

Regards,
Peter