Quite by chance I have just received an update of a review I did for [a gold open access scientific journal]. I omit all confidential info:
Dear Dr. Murray-Rust,
Thank you for your review of this manuscript. The Editor has made a decision on this paper and a copy of the decision letter can be found below.
You can also access your review comments and the decision letter by logging onto the Editorial Manager as a Reviewer.
[Dear Author… ]
Before your manuscript can be formally accepted, your files will be checked by the [publisher’s] production staff. Once they have completed these checks, they will return your manuscript to you so that you may attend to their requests and make any changes that you feel necessary.
To speed the publication of your paper you should look very closely at the PDF of your manuscript. You should consider this text to have the status of a production proof. Your paper will be tagged and laid out to produce professional PDF and online versions. However, the text you have supplied will be faithfully represented in your published manuscript exactly as you have supplied it.
So as far as the author and reviewer are concerned everything is driven by PDF (confirming Cameron’s experience). PDF is a well-known destroyer of semantic information. This, of course, is common to all publishers. We have allowed them to create this monster and force it on us.
PDF holds back the development of semantically supported science.