> In this process, deletion rather than expansion of the wording of the message is preferable, because if an ordinary message is paraphrased simply by expanding it along its original lines, an expert can easily reduce the paraphrased message to its lowest terms, and the resultant wording will be practically the original message.

This bit has me perplexed. If you had a single message that you wanted to send multiple times in different forms, wouldn't compressing the message exponentially limit possible variation whereas expanding it would exponentially increase it? If you had to send the same message more than a couple of times I'd expect to see accidental duplicates pretty quickly if everyone had been instructed to reduce the message size.

I guess the idea is that if the message has been reduced in two different ways then you have to have removed some information about the original, whereas that's not a guarantee with two different expansions. But what I don't understand is that even if you have a pair of messages, decrypt one, and manage to reconstruct the original message, isn't the other still encrypted expansion still different to the original message? How does that help you decrypt the second one if you don't know which parts of the encrypted message represent the differences?

It's mostly talking about the case where someone receives an encrypted message which is intended to later be published openly. If it was padded by adding stuff, an attacker can try to reconstruct the original plaintext by removing the flowery adjectives, whereas if things were deleted the attacker doesn't know what to add.