I'm working at a charity where we have moved from assorted W9x machines to mandriva. There is a W9x database application which has the peculiarity of allowing the user to create multiple identical field names per record, and thus has variable record lengths.

I'm being told that exporting this to csv and consolidating the data in multiple identically named fields will lose data.

(1) Is this plausible? I've done it, audited, and there is no doubt that every bit of data in every field has moved across. I cannot find any sort of grouping or sorting that can be done in the W9x app that I cannot do in Tellico. What could I be missing? The thing that is hard to see is, if the field labels are identical, how can you do things with the data split up into two, that you cannot do with them in one? Doesn't having multiple identical field labels limit rather than expand one's ability to manipulate the data?

Is this something about xml as opposed to csv in databases?

(2) Is it really possible that there are data structures which require multiple identical field labels and variable length records? Like, what and why? And could a collection database be one of them, and why?

(3) Is there a good source on open database format? The recent BECTA publication mandates odb or csv export for databases in the education sector. Where would you go to get a good account of odb? With particular reference to the moveability of a database from one platform to another.

Peter