The refereeing process, when done well, has great potential to improve your work and how well it is received by the scientific community. An important part of your role as an author is to guide the process toward this cooperative effort to improve the manuscript. The easiest way for the process to go off track is to think of the interaction with the referees as a debate, or even worse an argument, about the correctness and quality of the work, so make a particular effort to keep the process on the right track.

Structuring a reply to referees

Replies should aim to make it very clear and easy to see that you have comprehensively responded to every criticism from each referee.

Break down and quote all comments from each referee and generate a numbered list, as numbers will help everyone involved keep track of different issues. Use a different color or font to distinguish comments from the referee and your reply.
Make replies self-contained by quoting changes i.e., say “We have changed [original text] to [edited text]” or similar. Don’t write that you have made a change (somewhere) in the manuscript and don’t refer to an answer given to another referee, as the referee won’t want to search around for your response.

Things to keep in mind

Take care to ensure that you are directly responding to the question posed by the referee. If a long response is necessary or desirable, make sure that your opening sentence signals the main point of your response. Using the same terminology as the referee helps ensure a clear connection between the question and the response.
It is usually better to address the referee directly i.e. write “Thank you for raising this question” rather than “We thank the reviewer for raising this question”.
Try to respond to almost every point made by a referee with a change in the manuscript. Sometimes a referee’s comment may suggest a misapprehension on their part. Even if your initial feeling is that the referee has missed something that was already obvious, always treat this situation as an opportunity to make the text clearer and easier to read. Assume that the misapprehension happened because the manuscript either did not provide a sufficiently good explanation or that it did not properly emphasize the point in question and change the manuscript as best you can. This process almost always improves the manuscript and puts the referee in a much better position in terms of potentially reversing any negative opinions they have while saving face.
Consider using phrases like “thank you for the chance to clarify our text” as a means to convey the idea that the referee is reviewing an improved manuscript and to avoid undue emphasis on whether the original text was right or wrong.
Pick your battles. It is usually best to comply with linguistic or stylistic changes. If you get an annoying request to cite a marginally relevant paper it is usually best to do so.
It is standard practice to thank each reviewer, so follow this tradition. More importantly, it is often good to thank the reviewer in the body of the reply. This can be especially helpful if the report is overall negative or even outright nasty. Highlighting one constructive comment within a sea of negativity can help a lot in keeping the process cordial and constructive.
The review process often involves some degree of frustration. It is imperative to stay cordial in all circumstances. Bite your lips if you have to. It is often helpful to re-read what was written after a cooling-off period. Avoid CAPITALS and red text as ways to emphasize text as this can imply anger. Italics usually work best and bold is usually okay as an alternative.

Editor

Although most of the technical evaluation of the manuscript will be done by the referees, the editor holds the cards for the final decisions.

Keep in mind that they are usually very busy and that they almost certainly have a much better understanding of manuscript writing and the review process than you. They usually have a very good overall foundation in physics, even if they are not specialists in your area. Grandiose or verbose messages probably won’t get you very far.
Keep in mind that editors often have to reject a large fraction of submissions, so their tone will often be negative, and maybe more negative than their true feelings. They will, however, almost always be willing to read direct evidence-based arguments.
Do not make the mistake of assuming that the job of an editor is to count yea versus nay votes. In the case where a referee recommends publication but does not provide a good explanation why they are entitled to ignore this advice. In this case, it can sometimes be helpful to press the editor to obtain more reports that do contain reasoning behind their decisions, perhaps invoking the reputation of the journal. If reports directly contradict each other on a scientific matter this can make it self-evident that one referee is right, and one wrong, and one can use this to suggest the journal is not well served by taking the advice literally.