This is not a simple question at all, and not one that admits of a single solution.
To start with, here is my translation of SN 46.20:
Sattime, bhikkhave, bojjhaṅgā bhāvitā bahulīkatā ekantanibbidāya virāgāya nirodhāya upasamāya abhiññāya sambodhāya nibbānāya saṃvattanti.
“Mendicants, the seven awakening factors, when developed and cultivated, lead solely to disillusionment, fading away, cessation, peace, insight, awakening, and extinguishment. What seven?
Satisambojjhaṅgo … pe … upekkhāsambojjhaṅgo
The awakening factors of mindfulness, investigation of principles, energy, rapture, tranquility, immersion, and equanimity.
As you can see, in this instance, I have elected to spell out all the factors, while Ven Bodhi has abbreviated them in line with the source text.
This is a good illustration of the general policy I have followed:
- Reduce the number of abbreviations that refer outside the present sutta
- Abbreviate as much as reasonable within the present sutta
So the idea should be to minimize the amount of times a reader has to refer elsewhere to complete a passage. At the same time, we must acknowledge the fact that the texts were originally meant for oral recitation, and repetitions feel very different when reading as opposed to reciting. In my guidelines for translators I wrote:
Unlike most translators, I have been as much a part of the oral tradition as I have the written. I’ve been a patimokkha reciter for over twenty years, and in my youth I memorized dozens of suttas. I could recite Pali non-stop, from memory, all night if I wanted. Alas! It’s mostly forgotten now, though I still keep up the patimokkha.
After many hundreds of hours reciting Pali, it becomes obvious that the very same features that are most annoying in a written text often have exactly the opposite effect in recitation. The repetitions, the vocatives, all these things create a soothing rhythm. They give space and ease to the text, making it more enjoyable to recite. They also help to create “non-thinking” space, where you can contemplate the meaning as you recite.
But when reading, these same features become annoying. They’re a cognitive grit, constantly requiring filtering out. They encourage shallow skim reading, as you assume a text is not worth lingering over. While the linguistic content can be mirrored by imitating these conventions, the emotional response is exactly the opposite. And, as AN 5.26 teaches us, it is through a positive emotional response that a text becomes liberating.
However, general policies and practices aside, in such a large corpus there will be many exceptions and subjective choices to be made. Practically speaking, things get complex when you drift too far from the source text. Fortunately, the source Pali is usually quite reasonable in how it handles abbreviations.
The guiding consideration behind this policy is the fact that our text will be on the web. In this environment, it is no longer read from one sutta to the next, but is essentially “random access”. If reading a book, we can assume that someone reads one sutta, then reads the next, so they can remember what that passage said. Of course, even in books this is a big assumption and doesn’t always work, but still, it is true to some extent. Online it is still true to some extent, but less so. This is why I try to minimize having to leap from one sutta to the next. However, the practical reality is that in a general edition this cannot be totally eliminated. There are simply too many repetitions.
As to the ideas for offering brief and expanded versions, this is a tantalizing possibility, but it’s not as easy as it sounds. It’s not so hard if you’re just doing a few selected texts, but to do it consistently across many thousands of texts, most of which are abbreviated in one way of another, is a daunting task. That’s not to say that it can’t be done: we specialize in daunting tasks! But we haven’t done it so far, nor have we seriously discussed it.