Discussion List Archives

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Minutes - Action item (1) merging dictionaries

  • To: Multiple recipients of list <comcifs-l@iucr.org>
  • Subject: Re: Minutes - Action item (1) merging dictionaries
  • From: Brian McMahon <bm@iucr.org>
  • Date: Tue, 15 Feb 2000 12:01:19 GMT
Dear Gotzon 

> I have only a question about the "merging dictionaries" part. It is essential
> when creating a virtual dictionary that validates a given CIF the order in
> which the (real) dictionaries or fragments are concatenated and the form in 
> which name collisions are resolved. However although each CIF would include 
> the pointers to those dictionaries which should be used to validate it, there
> is no way of indicating how these dictionaries have to be concatenated. In the
> examples, dictionaries are merged and data files are validated by external 
> programs (cifdiccreate and dictcheck) that require some human intervention. 
> However validation of data files through virtual dictionaries could be an 
> automated task and therefore at least the mode of solving name collisions
> should appear also within the data file (assuming, for instance, APPEND for
> the dictionaries listed in the _audit_conform_ loop). Should some appropriate
> items be included in the core?
> After clarifying this point I would approve the document.

I confess that I hadn't really thought of specifying a virtual dictionary
for validation from within the CIF. I saw AUDIT_CONFORM as a category
whose main purpose was to list the public "standard" dictionaries used as 
a reference when the CIF was built. The freedom to layer other dictionary
fragments was designed to allow validation against other local criteria, and
so it made sense to make this a process driven by the validator. But I guess
that in principle the protocol should permit the CIF to carry an audit trail
of conformance against a specific virtual dictionary (which an end-validator
may over-ride if he wishes to impose local criteria). The recipe for that
virtual dictionary would indeed be carried in the AUDIT_CONFORM category. As
well as the method of overlay, you would need to specify the order in which the
dictionaries were loaded.

What do others think? If there is a consensus that such a facility is
desirable, I'll think about suitable data names. But if people are lukewarm,
we could defer defining new data names until a real need was demonstrated.

Best wishes