[cellml-discussion] Review solicited for application/cellml+xml

[Alan Garny]

Dear all,

Please find my 2 cents worth of comments regarding Andrew's Internet draft.

> We have several reasons for not encouraging anything other 
> than .xml as the file
> extension:
> 1) We have used .cml (which conflicted with other markup 
> languages, like Chemical Markup Language) in the past, and 
> changed to .xml. Having a history of too many file extensions 
> makes it hard to encourage software vendors to stick with a 
> particular one.

Just for accuracy, I believe to be the one who started using .cml in my
software (http://cor.physiol.ox.ac.uk/). Are there others who do so too now?

You mention that .cml conflicts with other markup languages, but I would
argue that .xml conflicts with even more markup languages. Personally, .cml
is not used on my system by anything else but for CellML files, so I don't
have that conflict you mention, while it wouldn't be the case if my CellML
files were to have the .xml extension.

As for a history of file extensions, again, as far I can recall (please
correct me if I am wrong), .xml has been the extension chosen for CellML
files from day one. It's just me who has been some kind of a dissident for
going for .cml from day one too (more on the reasons below).

> 2) There is currently repository software which uses .xml, 
> and changing this will likely create problems for (some) 
> client software using the repositories.

You will always have problems of some sort or another. To go from CellML 1.0
to 1.1 is not a trivial thing either, be it for the software developer or
modeller. To change the file extension from .xml to whatever surely is a
very trivial problem, at least of the kind that I would be glad to deal with
on a daily basis, if they were to be the only problems I had to have!

> 3) Many users of CellML like to edit their models as XML in a 
> text editor. Using the XML extension allows this to work on 
> systems which know nothing about CellML.

Many CellML users like to edit their models as XML in a text editor? Is that
really true? I certainly don't want to start an argument here, but I somehow
think that my users would feel very strongly about having to modify CellML
files using a simple text editor. I am, personally, very glad that I don't
have to use a text editor, nothing worse than modifying a CellML file than
using such an editor (very much error prone!). Allow me another quick plug
in: have a look at COR and you might change your mind and never use a text
editor in the future (though you might still want to do that once in a while
for whatever reason).

Also, it is very simple, on any operating system, to associate a particular
extension to a particular problem.

> 4) Using a specific extension encourages software vendors to 
> rely on this file extension. Although all software developed 
> at the Bioengineering Institute does not rely on a specific 
> file extension, there is software developed elsewhere that 
> will not open files which do not have a .cml file extension 
> (even when the user specifically tries to open that file in 
> the program), and I have heard that this is creating problems 
> for users. 

I believe that it would have to be my software indeed... I have had several
email exchanges with David Nickerson about this over the last few years. In
a nutshell, the main reason I am rather reluctant about the idea of using
.xml for CellML files is that by having using .cml, I can associate that
extension to COR, while the same cannot really be done with .xml.

My point is that COR has been designed from the beginning to be as user
friendly as possible. It is something that is very high on our agenda and
that has surely been very much appreciated by our users (from what I
know/understand, there is no way they would use a text editor to edit CellML
files for instance).

To make COR user friendly involved associating .cml files to COR, because
all the users have to do to open a CellML file is, for instance, to double
click on it rather than opening a program, then opening a CellML file. It
may not be too much to do the latter, but why not make the user's life
easier if all that takes is associate an extension to a particular program?

> By having only a more generic file extension, we 
> encourage software developers not to use the file extension 
> to determine whether they should open a certain file, and 
> instead look at the MIME type (where available) and/or the 
> XML namespace.

Yes, that could easily be done, but... :)

	Best regards, Alan.

--
University of Oxford, Department of Physiology, Anatomy and Genetics
Sherrington Building, Parks Road, Oxford, OX1 3PT, England
http://noble.physiol.ox.ac.uk/people/agarny/