Peter Gulutzan

Milestone 1.8.3 should include these "Top Five" matters: 1. Data Types -- Column definition in SQL #2351 2. Identifiers-- Identifiers in SQL #2345 3. ICU -- Integrate ICU into Tarantool....

A redundant grant should not cause an error, so, if there is an underlying box.schema.user.grant request, it should contain {if_not_exists=true}.

I think we disagree about what the priority is. I said that FOR EACH ROW should not be optional, that is, it should be a mandatory clause. That's simple, eh?...

In issue #2579 I commented that it's bad when there are no rows, because that causes an error message. Would a BEGIN step solve this by letting users initialize the...

Also, if I want to use AUTOINCREMENT, the only thing that seems to work is: box.schema.user.grant('guest','create','universe')

I certainly never would have suggested "w/o Lua at all". SQL's Unicode escaping (inside U&"..." and U&'...') would not conflict with Lua escaping. I think the issue here is: should...

I'll interpret Konstantin's remark as "yes, invalid UTF-8 characters should be legal in text/char". This means that one of the reasons that we would distinguish text/char from blob, is no...

I think this is obsolete. See Issue#5809 Segmentation fault or error message if bitset index exists

If we were trying to be compatible with PostgreSQL and MySQL, "local" might be important. But if we're trying to be compatible with standard SQL and the rest of the...

Summary of work involved: * Persuade everyone to acknowledge that session-local temporary table definitions are not required, but session-local temporary table data rows are required. * Require the word GLOBAL,...