Re: [Psidev-ms-dev] Difference between purgatory / obsolete

[Matthew C. <mat...@va...>]

We used purgatory as a development limbo. We may be at a sufficient 
stage of release where we can delete the purgatory terms entirely (which 
we will never do to properly obsoleted terms). I don't think purgatory 
and obsoletion should be considered overlapping concepts. I don't recall 
that there should be any mzML documents at or after 1.0 which use terms 
that are now in purgatory. I'd say a term that was NOT in purgatory in 
the 1.0 mapping file may be obsolete now, but it shouldn't be in purgatory.

I think of supporting obsoletion like deprecation in a programming API. 
At some point in the future, support for those documents may be stopped. 
Perhaps we should have at least one converter which could upgrade the 
documents to replace the obsolete terms if the term has an official 
"replaced by" comment (or tag) and to delete the ones that were 
eliminated entirely. Or possibly handle those on a case by case basis 
(like when isolation window semantics changed from "lower bound/upper 
bound" to "target m/z/lower offset/upper offset").

-Matt


On 4/8/2010 3:32 AM, Steffen Neumann wrote:
> Hi,
>
> I didn't come across a halfway formal definition / distinction between
> OBSOLETE'ing stuff, and moving it to purgatory, although we had
> a thread on this here last year ("Marking CV terms obsolete")
>
> Also I'd expect some process to make sure that stuff
> shoved into purgatory really should be obsoleted at some stage.
>
> This has also implications on the validity of instance documents,
> because files which have some cvParam which IS_A "something"
> (where "something" is a MUST) are invalid once the cvParam
> is moved from "something" to "purgatory", *unless it also retains*
> the original IS_A "something".
>
> I found (grep -B 1 -A 1 "MS:1000479 ! purgatory" psi-ms.obo)
> that there are Terms with a mix of:
>
> 	relationship: part_of MS:1000479 ! purgatory
> 	is_obsolete: true
>
> 	relationship: part_of MS:1000479 ! purgatory
> 	without is_obsolete
>
> 	is_a: MS:1000479 ! purgatory
> 	without is_obsolete
>
> So:
>
> 1) Why do we use purgatory in first place ?
>
> 2) Do we want a policy whether and how the CV is cleansed
>     and purgatory stuff marked OBSOLETE ?
>
> 3) How do we handle documents which become invalid
>     with these changes ?
>
> 4) Should these definitions/processes
>     become part of the spec documents ?
>
> Yours,
> Steffen
>
>