Re: [ddlm-group] Relationship asmong CIF2, STAR,CIF1 and Python. . . .

On Tuesday, January 18, 2011 7:20 AM, Herbert J. Bernstein wrote:

  Now I am very confused.  You say we have not broken the promise on the
IUCr web site, but at the same time we seem to be defining a CIF2 that
will not accept CIF1 documents.

  Please bear with me, and, even if you think it has already been
explained, please explain precisely how to use CIF1 documents in the
currently proposed CIF2 environment.

  If we have a sound way in which a CIF1 document has use of a DDLm
dictionary, then we do not need to bother most of the community with CIF2
for data files at this time.  All they need right now is what I called
DDLm-2011, a CIF2ish DDLm dictionary format.

I agree with that assessment of need, but I don't see what would be gained by limiting CIF2 release like that.  If CIF2 is not ready or appropriate for data files, then I think a CIF2-like DDLm-2011 language leads users and especially developers in the wrong direction.  If we wish to release DDLm without unleashing CIF2 on the world then let the initial DDLm and dictionary releases be crafted in an altogether different format, such as XML.  In the unlikely event that there were genuine interest in such a course, it would be worth mentioning that I have a suitable XML schema at hand, as well as supporting software that could easily be adapted to translating existing DDL and dictionary documents.

 If we don't have a sound way
in which a CIF1 document has use of a DDLm dictionary, then I think we are
breaking the promise on the IUCr web page.  Please recall that DDL2
dictionaries are not valid CIF1 documents -- they have save frames, so it
is not unprecedented to have a different spec for dictionaries as opposed
to data files.

I accept that, but it's a different matter for the data format to be a subset of the dictionary format than for the data format to be a related but subtly incompatible format.  We will have that anyway when DDLm dictionaries are used to validate CIF 1 files, bet let's please not set it as the direction for the indefinite future.

 It makes a big difference to most of the user community if
we are simply telling them we have a new dictionary format rather than
telling them we are changing the data file format.

Agreed, in that much of the user community doesn't care about dictionaries.  On the other hand, members of the user community who care about some of the new CIF2 features -- Unicode support, as a prime example -- would not necessarily take the distinction as a positive or even a neutral proposition.

  On David's description, I think I really did explain why I think we will
have trouble populating missing values involving CIF1 tags that are not
valid CIF2 tags.  Doing that using the alias mechanism would seem to
require defining the CIF1 tag in the DDLm dictionary as a primary
definition and then aliasing a CIF2 tag to that primary CIF1 tag, so that
a method working with the CIF2 tag would effectively populate instances of
the CIF1 tag, but, and this is the part I can't seem to get past, defining
the CIF1 tag in a new CIF2-style DDLm dictionary would seem to require
that the CIF1 tag be a valid CIF2 tag.

I think we will not easily get past this dispute without an example.  For that purpose, then, perhaps James, David, or another participant with practical DDLm and dREL experience would be kind enough to present a solution to this exercise:
Provide DDLm definitions and a dREL method that support computing a missing value for the Core item _exptl_crystal_density_diffrn, based on Core items _chemical_formula_weight, _cell_formula_units_Z, and _cell_volume.  The definitions presented should use DDLm formalism for the defined data names, and should be compatible also with validating the corresponding mmCIF data names.

James's and David's comments have given me every reason to believe that this would be straightforward, though the definitions together with their required context might be bulky.  I am hoping that the requested definitions are in fact already written.

 I suspect we will get into trouble
in other areas of using existing CIF1 tags in CIF2 DDLm dictionaries.

One of the key promises of DDLm, as I see it, is that the distinctions between various syntax versions and between DDL1 and DDL2 formalisms are relevant to only two program activities:
1) On input, reading a file correctly and associating data items with the correct DDLm definitions.
2) On output, producing well-formed files for the target syntax version that are valid with respect to the DDL1 (or DDL2) dictionaries with which the DDLm dictionary provides compatibility.

As long as those two features work correctly, details of syntax version and original target dictionary can be completely abstracted away from validation and dREL operations, leaving no room for other areas of trouble.  Success in those areas will be a function of program, DDL, and dictionary details.  CIF2 syntax need only be sufficient to support the required DDLm features; it does not otherwise bear on the problem.

How important each of those trouble may be depends on our goals, so I
respectfully urge that we make certain that we are working from common
goals, so that we can then focus on whether we are meeting those goals,
rather than have debates that seem to be based on different goals for
different speakers.

That is a reasonable criticism of our process to date.  I am willing to participate in the proposed goal re-evaluation process, and I hope it will help resolve some of our current disputes.  Of late, however, we have also seen significant differences in technical analyses that should be independent of participants' goals.  Therefore, I do not anticipate that the goal re-evaluation exercise will provide clear resolutions to *all* our current disputes.


Regards,

John

--
John C. Bollinger, Ph.D.
Department of Structural Biology
St. Jude Children's Research Hospital


Email Disclaimer:  www.stjude.org/emaildisclaimer

_______________________________________________
ddlm-group mailing list
ddlm-group@iucr.org
http://scripts.iucr.org/mailman/listinfo/ddlm-group