fenway

This is a pointless argument... you can't JOIN a permission from inside a SET.

No, I want you to add those as ON conditions.

The most "general" is to have full control over the permission records -- which requries them to be records.

dating_user_info info ON sessions.user_id = info.user_id Add to the ON clause.

There's a missing comma after t.id.

Well, at least for the INSERTs, you can probably use DELAYED / IGNORE. For the UPDATEs, you can use LOW_PRIORITY.

ALTER TABLE customfields MODIFY COLUMN `cust_field_type` varchar(100) NOT NULL default '';

See here -- it will log everything.

Post this live somewhere

Don't know cpanel... but you can issue a simple ALTER TABLE statement.

1M users? You'll have other issues by then.

Hmmm... you can try adding the range limits to the ON clause in the JOIN with the info table....

As I mentioned before, my preference is a separate permissions table... if it's indexed by user_id, the cardinality is good, and it should be relatively fast to get back a few permission records; then it gives you room to grow.

Why? Run the fast query, then limit the results in PHP.

What's wrong?

Assuming you have your own DB wrapper class, this is simple... otherwise, there's a query log, but it's binary...

Depends on how many indexes have to be updated...

What's the query? Don't blame an entire RDMBS for a syntax error.

You can't index a bitmask the way you're thinking.

Using flat files is just asking for trouble -- there's a reason they invented databases.

Check out LOAD DATA INFILE in the refman -- your delimiters can be just about anything.

There's no way to query across an entire DB with a single query -- you'd have to look into 3rd party indexing solutions.

You want to do what?

Your grants table doesn't allow connections for that user from that ip.

Subquery != dervied table -- at least if it's not correlated. Subqueries can be bad because they run for each row -- long as it's just run once per query, it's not as bad. Yes, there is no indexability; if you're really concerned about this, you can create a temporary table in memory, index it, and then join it properly.