Hi Andrea, Bernard, et al,
I've been sitting on my hands to see if there was any reaction to the way UCDs are heading, but as I was mentioned in the last post, I feel I can free my hands and re-phrase my concerns again.
I fully agree with Bernard's position, which something isn't quite right with the way UCD1+ are heading, particularly in the photometry section, which comprises and will comprise the bulk of the data in astronomy.
We should ask ourselves why we want to attach a UCD to a column's metadata, and how we intend to use it.
To the first question, the initial answer (pre-historic UCDs :-) was that we wanted to be able to recognize the content of any given column in as accurate a way as possible in order to compare it with others (we had datamining in mind); we wished not to mix apples, pears, oranges or lemmons, so, there would be different UCDs for apples, lemmons and oranges (to follow Roy's way of exemplifying things) :-)
Now, we have reached a point in which we're lucky to have one ucd for
apples and pears and one for lemmons and oranges (citrics). Perhaps we
only have one UCD for
fruits.which.grow.on.trees.of.more.than.two.metres.high :-)
Which leads us to the second question... How do we intend to use them?
Do we want to find "things" using ucds? Sure,
1- I(*) want to find all catalogues containing observations with the
gunn g filter (enough of fruits, we're astronomers :-). Perhaps I'd like to compare observations done by different observers?
(*) replace I by "user" or "users".
2- Another possibility is that I want to find observations in the optical
blue section of the em-spectrum? Fine, I may be just interested in discovering resources rather than comparing on a very accurate level.
3- I want to discover which objects have been observed in the radio regime
of the CO emission at 3.whatever mm
[go back to normal use of 'I']
As far as I can see, ucd1+ allows us to answer [2] accurately and possibly completely, while [1] and [3] will give us a lot of answers with a low S/N, ie, correct answers combined with irrelevant information (to the person questioning).
One of the initial ideas of UCDs (mine at least) was to tag columns accurately (including filter/photometric systems), but having "virtual UCDs", ie, UCDs which will be accessible to those querying, but never (or rarely) attached to columns. Translation tables would take care of interpreting that if I asked for "blue magnitude" I would get Johnsons' B, Stroemgrem, gunn, washington, HST, Sloan, etc, etc. However accurate querying was possible.
The problem of UCD1+ is that if I want HST Blue observations I have to weed through all the other systems which were put in the same box!!
Somehow I feel as if I went to a supermarket to buy AAA alkaline batteries and I'm told "Ah! batteries, they are all in this box (AAA, AA, C, D, N, etc., alkaline, zn, li, new-tech, etc.), here, help yourself" I don't think I'll go back to buy batteries there again!
OK, so what can be done? I'm not just complaining here! :-)
UCD1+ have already the broad classification, *I* think we need to add the fine structure, so no information is lost and needs to be re-discovered time after time
For instance, S|em.opt.B should be left for a magnitude for which we have no idea how it was measured (old catalogues), while a Johnson's B should be something like S|em.opt.B:phot.jhn.B
Yes, the ucd tree will grow again (or we'll have to handle this with UTYPE or some other meta-data element), but the example just given gives an accurate description of the quantity, yet can be used in fuzzy searches.
Alternatively, we could have phot.jhn.B by itself as the UCD, but have something like
S|em.opt.B -> (S|em.opt.B, phot.jhn.B, phot.HST.B, ....... )
That means that in the context of a query, whenever a tool finds the left term it should look for columns which contain a UCD listed in the right hand side. If no list is associated with a "virtual UCD", exact matches should be performed (which should not prevent us from searching for 'phot.HST' regarding of the filter observed! I've implemented tools like this, that's why I'm voicing this proposition.
To finish, I just took a quick look at the 'em.line' section... well, absorption and emission lines exist across the spectrum: Ly-alpha, CIV (QSOs), molecular lines, (CO, H20) in Radio, FeXIV in X, etc. By not leaving room for accurate description with UCDs we are making UCDs less useful as accurate descriptors and discovery tools.
I feel more threatened by a lack of accuracy in the description of a column than for a largish set of terms (which only reflects our trade).
Cheers,
Patricio
On Mon, 9 May 2005, Bernard Debray wrote:
> Dear UCD list members,
>
> I apologize if the subject of the remarks below has already been
> addressed or this turns out eventually to be outside the scope of the
> UCD field. This is both a remark on the UCD electromagnetic spectrum
> scheme and a question about the use of UCDs.
>
> I have browsed the archive of the UCD mailing list and found only one
> message, by Patricio Ortiz, already a long time ago (October 2003,
> http://www.ivoa.net/forum/ucd/0310/0067.htm) that fits this topic and
> looked quite sound to me.
>
> One of the claimed aims of UCDs is interoperability. Therefore, they
> should serve, for instance, the purpose of comparing magnitudes from
> different catalogues in given filters. Therefore, one should be able to
> identify unambiguously through which filter, photometric system, ... a
> magnitude is supplied in a catalogue or by a simulation tool.
>
> In the division of the electromagnetic spectrum in the UCD document,
> optical and near-infrared subparts of the spectrum where given names
> which are the ones of filters in the Johnson, ... photometric system,
> whereas they designate contiguous portions of the electromagnetic spectrum.
>
> While Johnson filters do fit in the subparts so defined, this is not
> always the case for filters from other photometric systems. For
> instance, the g filter of the SDSS photometric system has an effective
> wavelength of 482.5 nm but the flux in this band encompasses widely both
> the em.opt.B and em.opt.V domains.
> On the other hand, there seems to be,as far as I can see, no UCD which
> refers to the specification of a photometric system which could
> complement the UCD for the description of the electromagnetic spectrum
> subpart.
>
> Alternatively, the UCDs for specific filters such as those of SDSS'
> system, which could become more widespread that Johnson's in the future,
> could be handed over to particular namespaces. But it then still seems
> to me that, in order to release ambiguity in the meaning of optical and
> near-infrared parts of the spectrum, the corresponding UCDs should be
> named slightly differently, not simply U, B, V, ...
>
> Finally, if the desired achievement of identifications of data columns
> in the purpose of comparaison is beyond the scope of UCDs, should the
> cross-comparaison process be handed over to VO tools at more
> sophisticated levels ?
>
> Thanks in advance for your remarks and comments.
>
> Best regards,
> Bernard Debray
>
> Bernard Debray bernard.debray-at-obs-besancon.fr
> Observatoire de Besançon http://www.obs-besancon.fr/
> BP 1615
> 25010 Besançon Cedex
> France
>
>
> Andrea Preite Martinez wrote:
>
> >Dear member of the UCD_wg,
> >
> >since last Friday (20050429) an updated list of UCD words is available at:
> >
> >http://vizier.u-strasbg.fr/UCD/ucd1p-words.txt
> >
> >at the UCD page:
> >
> >http://vizier.u-strasbg.fr/UCD/
> >
> >The new list of words has been edited according to the very valuable
> >comments coming from the community and also thanks to the (almost!) every
> >day use of them made by Sebastien and myself.
> >
> >The most important differences with the previous version (20040823) concern
> >
> >- distances and sizes (linear/angular)
> >- the introduction of a rectangular/cartesian reference frame
> >- the phot.color section, much simplified
> >- spelling, abbreviations, syntax ...
> >
> >
> >At the UCD page you can also find un updated version of the UCD_tree
> >browser(s)
> >and a new tool, the ucd_builder:
> >
> >http://vizier.u-strasbg.fr/UCD/cgi-bin/descr2ucd
> >
> >that uses a description in "natural" language to find the relevant
> >ucd_words, and then tries to build a valid ucd out of them.
> >
> >In these two weeks before Kyoto please explore the list, use the tools and
> >send back comments.
> >A more "formal" comment page is available to suggest changes in the list of
> >words at:
> >http://vizier.u-strasbg.fr/UCD/cgi-bin/comment/ucdComments
> >
> >Have fun.
> >Andrea
> >
> >
>
>
--- Patricio F. Ortiz pfo-at-star.le.ac.uk Department of Physics & Astronomy Phone: +44 (0)116 252 2015 University of Leicester Leicester, LE1 7RH, UKReceived on 2005-05-09Z20:32:17