Clive Page write
...
> Actually a good way of setting up a "standard" is for two (or more)
> places to agree something and interchange data using it. That's how
> FITS started. These GSFC/HEASARC conventions are now quite widely used.
I agre with most of what Clive has said. I would also like us to remember that all these data formats which we are finding drawbacks to (VOTable, FITS etc.) were supposed to be standard and everyone thinks that their version either is standard, or should be setting the standard. We can dream up what seems like the perfect replacement but astronomers will find a way to hack it because it doesn't quite do what they want. A large amount of this is uneccessary and happens because there isn't proper user-friendly documentation, but quite a lot happens because e.g. unforseen accuracy becomes possible.
The trick is not to try and solve all possible problems before they arise, that's impossible; the trick is to realise that standards evolve and make sure the formats start simple, get tested in _real_ use and can be revised. Or have translation layers provided. VOTable seems relatively easy to evolve; FITS requires metadata or header translation for non-standard flavours which can be done easily, and I have yet to see any remotely feasible replacement format for massive binary datasets, especially given the huge amount of software which expects FITS.
In other words, whatever the VO produces, astronomers will break it, we should be concentrating on getting things which can be used now so that we can find out what is most needed to be flexible and easy to heal...
cheers
a