Hi Oleksandr, I'm going to stop bitching about Maria now. My needs are in fact met by the current dyncol functions: - It's not so hard to write methods that generate CREATE_COLUMN() expressions from a general data structure. - COLUMN_JSON() is just fine for all my reading purposes. I see no need for a naked COLUMN_GET() in a SELECT. - Generating COLUMN_GET() expressions for use in queries ends up being a regex replace with some packaging. - As I said in the email to Frederico this morning, indexes are not unimportant but a lot of apps can live without and many will be fine. When I wrote my original email, that list of seven concerns were factors in my vague worry that dyncols are going to remain ignored and unloved. That original email is adequately answered and I have the information I need. Tom