[moby] Re: [MOBY-dev] how to handle errors in a biomoby service?

[David González Pisano]

Hello,

Mark Wilkinson escribió:

>Hi all,
>
>Thanks for writing this up so clearly!  I agree 100% that the current
>error-reporting system in BioMOBY is insufficient (it was never part of
>the initial proposal, and even that initial proposal was cut by 80%,
>so...)
>
>I am a bit wary of two aspects of the solution you propose, though I
>need more time to think about exactly why they make me nervous:
>
>1)  Mixing metadata (errors) into the data seems... well... wrong
>somehow.  I would prefer a solution that somehow used the serviceNotes
>block as the place to put error messages, and added an attribute to teh
>mobyData, Collection, or Simple tags that Xref'ed a node in the
>serviceNotes block.  I wasn't in Malaga for the discussion, so perhaps
>there is a strong argument for putting it there that I am unaware of?
>
>  
>
There were 2 proposals in the document and yes, the second one had the 
errors mixed with the data. We will write it again to clearly separate 
the moby data from the error information (like in the first proposal, 
that uses serviceNotes) and crossref them someway (see below)

>2)  overloading the articleName attribute in the Simple/Collection tag
>(and especially in the case where Simples are part of a Collection)
>would then give **three** different "meanings" to the articleName
>attribute!!  It's bad enough that we already have two different uses for
>that attribute... I'm reluctant to add one more to the fray! :-)  Let's
>create a new attribute instead of adding a new meaning to an existing
>one.
>
>  
>
We understand that the articleName attribute in Simple / Collection tags 
is the "anchor point" we have to use to refer to that Simple / 
Collection. Referring to Simples / Collections / Simples into collection 
using articleNames give us more granularity than referring to mobyDatas 
using queryIDs. There is no need to create another additional 
"identifying" attribute if we can identify the entity we are reporting 
an error for. Probably we have to rewrite that part to clarify several 
points:

- The error tag has to use an attribute to refer to the corresponding 
failing simple / collection. We can call this attribute exceptionRef, 
articleNameRef or something similar, to avoid using articleName again in 
a different context, but
- We are referring to the erroneus entity using its articleName, and 
that's why the articleName has to be mandatory (note that this is a 
significant change in the bioMOBY specification, but we can't figure out 
any other way to report errors at this level)

David

>I hope others will participate in this discussion - it's been pretty
>quiet out there!  I have not thought deeply about error reporting, and I
>would prefer to leave the final decision to people who *have* thought
>deeply about it.  I am perfectly happy to simply sign-off on whatever
>the expert group decides - I really do not want to be the arbiter of the
>final decision in this matter because I just don't have the full scope
>of the problem in my head!  I am happy, however, to participate in the
>overall debate leading up to that decision.
>
>M
>
>
>
>  
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dgpisano.vcf
Type: text/x-vcard
Size: 338 bytes
Desc: not available
URL: <http://lists.open-bio.org/pipermail/moby-dev/attachments/20050817/91cab6be/attachment-0001.vcf>