[ska]

This is correct, although published is pretty broad it doesn't include everything under the sun.

The real problem is that the patent examiners are not in general required to look at the whole breadth of what is published, just what is in the patent databases.

As a result there can be clear prior art, but unless it gets fed in during the review process, the practical way to resolve that is likely to end up in court, and likely to be expensive.

Of course often "clear prior art" isn't too. Hence the process to sort it out.

[AlbertCory]

Interesting side note about non-patent prior art (by the way, the examiners do, most of the time, look for it):

I was told that a proposal to require the examiners to use Google was rejected by the union: if you're going to require us to do more work, then you have to pay us more.

This was back in 2012 or so. I have no idea if now they have to use Google, but they only get about a day and a half on each patent (again, this is old data).

[ska]

> (by the way, the examiners do, most of the time, look for it)

Sure, but usually superficially (and not, I think, required); which is fair - you can't practically require the patent examiners to have the kind of breadth and depth necessary to have a good feel for this.

[AlbertCory]

One more point about this: the date on the publication isn't necessary the effective date. If it's a journal that says "May 1999" that doesn't mean the date is May 1. It would be the date that it became available to persons in the art.

Getting a librarian to say "yes, we cataloged it and made it available on May 5" would be pretty strong. Or a Wayback Machine archive from May 2.