Pete,
Requesting DCX means that your application is "unpredictability
resistant". In fact it asks for it. Our applications can be smart enough
to deal with the DCX type of unpredictability. It is very important
though to know which aspects are unpredicatable. I do not mind having an
unexpected extra field in a record; our applications can ignore those. I
do mind for example records that have a author field of 25 Mb or an
author name in the wrong field.
I think that metadata formats are used to ensure some aspects or some
rules of metadata, but some unpredicatability is unavoidable in order to
enable evolution.
Theo
>>> [log in to unmask] 9/17/03 12:54:37 nm >>>
Theo,
> Yes what you describe is the expected behaviour. However
> provider 5 may choose what is best ABC or XYZ or a superset.
> The dataprovider knows what is the best schema. One of the
> advantages of DCX is that providers do not have to worry
> about creating extra schemas.
Thanks. I just wanted to be clear that I had understood your proposal
correctly.
> I would expect the following table.
>
> request - possible response
> DC - DC
> DC - DCX (it is up to the service provider to return an error
> message to the user)
> DCX - DCX
> DCX - DC
> DCX - ABC or XYZ and the requestor prays for getting at
> least some
> DC properties
> ABC - DCX (it is up to the service provider to return an error
> message to the user)
> ABC - ABC (normal situation)
> ABC - DC (it is up to the service provider to return an error
> message to the user)
> ABC - error message "we do not support ABC
> DCX - error message we do not suport DCX
> DC - error message we do not support DC
But in the specific context of OAI-PMH, the only mechanism a service
provider has to request a specific "type" of metadata record is using
what OAI-PMH calls a "metadata format".
And if as a service provider I issue a GetRecord request for a record
for item oai:myrepos:maths/12345 using metadata format X, assuming
that
item exists, the repository has to return a GetRecordResponse with
either
- a metadata record for that item in format X; or
- a "cannotDisseminateFormat" error condition
i.e. if I request format DCX, and the repository returns a metadata
record, it has to return format DCX (not format DC or ABC or XYZ). It
may be that the content of the DCX record turns out to be the same as
if
I requested one of those other formats, but it is DCX format.
I think my point was that using a metadata format based on DCX in
OAI-PMH means that the service provider has to expect unpredictable
record content.
Of course, I can see that from the perspective of your proposal this
is
a Good Thing! But it does seem perhaps slightly contrary to the way
metadata formats have typically been used in OAI-PMH - though I say
this
without first-hand experience of working either as a service provide
or
data provider so I may be skating on thin ice, and others are better
placed to comment. My understanding is that metadata formats are used
precisely to try to ensure "predictability" - if as a data provider I
say I provide a specified format, then a service provider knows
_exactly_ what to expect.
(Having said that, I take the point made by Roland that as terms are
added to the DCMI dcterms vocabulary, "Qualified DC" as described in
Andy's model doc changes through time, and so would a metadata format
corresponding to that.)
Pete
|