A relatively recent acquaintance has inspired me to take up the pen again at cotidie, and write daily in Gaelic, Latin, and Greek (one a day, alternating). Since I'm generally good at pontificating about things I know better in theory than practice, I wanted to jot down some thoughts on the matter.
1. To write, you must read; but reading won't make you a writer.
I've read in my relatively short time not a small amount of Greek and Latin texts. And it's made me an excellent reader - I can generally read the NT without much pause or internal translation, and I can manage with not overly complicated Attic and post-NT Koine. Likewise I can read Latin with some fluency, though poetry tends to hold me up. But neither of these translate to quick and easy writing. There have been times where I've been quite speedy at Greek writing, but that has receded.
Nonetheless, one you have a practice of writing, reading is essential. Because as you continue to read you notice things: stylistic features, syntactical patterns, idioms, words you were looking for, and you incorporate them back into your writing. This is a far more useful process than constantly monitoring your own writing and looking everything up for technical answers. Write freely, with what you can summon in your head, and whatever mistakes come, and leave most of the self-correction for another time.
2. Accents and vowel lengths can wait (but not forever).
The trouble with these two features is that you really can ignore them 98% of the time for reading comprehension purposes. That makes them seem like unintegral features of the language, but they aren't. They mark semantically significant things, and so you should learn them. And yet when it comes to writing, I think it's best just to mark what you do know, and leave of what you don't, and again just add more correct detail as you improve.
3. We really need good English->Target dictionaries.
Traupman's Latin-English/English-Latin is the most useful for modern composition; Whittaker's words is the most useful Latin computer program ever devised, only hampered by its lack of vowel lengths. What I would love to see it a fairly comprehensive database/program/dictionary that gave full target language details (word, with accents/vowel lengths, with their lexical forms not just the head word, plus some words to explain nuance and usage. This is true even more so for Greek, which is woefully underresourced, in my opinion. Apart from Woodhouse, one can hunt through various vocabularies from composition books, but it's not a great situation.
4. What if we skyped? Structure and chaos in the blind leading the blind.
Two years ago I tried meeting up with one man to do some conversationl Greek, it didn't go excellently (as I'm sure he would acknowledge). I think we lacked structure and I suspect the internet gives the impression that I'm better at this than I am. There are a few people on my skype list who are there really for Latin conversation, but I never find that I sit down and think, "wouldn't mind a chat in Latin".
I think these things could work with more structure. In may Gaelic class, for example, we practice three kinds of exercise: Two translation exercises where a partner reads in either Gaelic or English and the other partner gives a translation in the other language. These are scripted, but the response is meant to be done without looking. We start with Gaelic->English and then later use the same sentences for English-Gaelic. Is this pedagogically ideal? Probably not, but it's a step more and better than nothing. The other exercise we do is conversations with a number of 'starter' questions from the week's material.
It makes me wonder whether such things couldn't be written up and used for Latin, Greek, etc.. Perhaps I will try and incorporate them into my slow moving FSI conversions.