Re: [Psidev-ms-dev] mzML 0.99.0 submitted to document process

[Marc S. <st...@in...>]

Hi all,

first of all i would like to thank Eric and all the others in the 
working group for their effort.
Here are my comments:

(1) The new CV term problem
A is clear and simple.
B is simply a bad idea in my opinion. Why not use the child accession if 
we have it?
C helps the software to know where the new term belongs, but the 
software does not know what to do with it in most cases. I think most of 
software implements these enum-like CV terms as enum types and thus 
cannot handle new values anyway. Additionally it is error prone 
(mismatching parent and child).

As C is an extension of A, i vote for A or C, but i don't think that C 
helps very much.

(2) Semantic validator
The semantic validator is a nice feature, but i think you must publish a 
file that defines the mapping of CV terms to the schema.
This file must answer questions like: Where can i use which term? How 
often can i repeat a term? etc.
With the heavy use of CV terms such a file is a non-optional part of the 
format definition.
What happened to that format Luisa proposed?

(3) Comments to CV / Schema
- The term MS:1000543 "data processing action" is missing some child 
terms i think. What about smoothing, baseline reduction and  removal low 
intensity data points?
- Putting the software name in a CV will cause much trouble i think. 
Where are way to many upcoming tools and you will be constantly updating 
that obo file. I really think we should put that into a string attribute
- I would add a new optional and unbounded element "parameter" with 
attributes "name", "type", value" to the dx:dataProcessing element to 
store the parameters of the software that were used for processing.

(4) General
Finally i'd like to say that i agree with Brian Pratt. There is too much 
CV and too little XML in the format for my taste.
I don't argue against CV in general it's a nice technique that allows 
the schema to be stable for a long time.
But now everything is in the CV and there are hardly any XML attributes 
left. This makes the format hard to implement and impossible to check 
with an XML validator.
And i don't see the advantage in most cases: I have to adapt the 
software to new terms just as i would adapt it to new XML elements.

Best regards,
  Marc