If they thought the analogy of tables and databases was a bad one, why did it take them like a thousand releases to know this.
Because you learn from your community and customers. And this could take time to make a decision.
Architecturally speaking this is so poor not to have thought of this earlier.
What would have been the consequence of thinking of that one or two years before? Same "pain" for existing users I'm afraid.
But as far as I recall we have been advertising that a lot the last years that using multiple types should be avoided. At least at conferences, at our trainings and on this forum.
When was the last time in an established db like oracle...
Oracle has been there for 40 years I think. And indeed this is pretty much stable now.
Note that you can still use the same "type" model you were using before the change. As explained in this blog post: Removal of mapping types in Elasticsearch 6.0 | Elastic Blog
For example, if you had:
POST index/type1
{
"foo": "bar",
"type1": {
"a": "text"
}
}
POST index/type2
{
"foo": "bar",
"type2": {
"a": 1234
}
}
You can now index it as:
POST index/_doc
{
"type": "type1",
"foo": "bar",
"type1": {
"a": "text"
}
}
POST index/_doc
{
"type": "type2",
"foo": "bar",
"type2": {
"a": 1234
}
}
And you will have a similar behavior as what you had. But now you can understand exactly the consequences of having multiple types per index.
This is absolutely ridiculous!!
Even though I understand your frustration, I don't think it's ridiculous. It has been a hard decision to take to be honest but sometimes you must take decisions instead of letting users shooting a bullet in foot. I'm sorry you have to suffer from it and we will be happy to help you on this forum to move forward.