I've tried space. About sharing: how about having unique textual IDs (say, UUIDs) for each space entry, that get carried around in cards and used to update cards in place?
Then space entries.spc deckname would update the entries in deckname, but also invent and add such IDs to any entries in entries.spc that are missing them. Then all we need to take care with is to only check in spc files that have IDs for all entries, which also serves to ensure the spc file is used at least once before checking in.
Since the ID is per space entry, not per card, need to figure out how to deal with intervals. Inheriting the smallest interval in any related old cards for all new cards seems just-workable, but not elegant at all.
But I think the proposal of this page is great, and update would be necessary, so why not. The other approach that might be better would be to change anki itself to use textual sources natively, but that would probably be a much larger change.
The problem is that the Space file format is intended to be human-readable plain-text. I'm not willing to compromise on that personally; I want to have plain-text files that I can check into a repository.
Without this, I think the best thing you could do is fuzzy matching on the content, but I'm not sure how well this would work in practice.
That said, you could adapt the markup language for use in Anki's editor, storing it in something structured like XML or JSON that would have UUIDs of the type of propose.
Also, thanks for trying space! It's not really polished or even finished and portable, but it's really gratifying to see people touching things I've made.
Spaced repetition is optimal for recalling factual information. It won't necessarily teach you anything that you haven't already learned. It helps you retain knowledge, and won't necessarily help you develop skills. But, within the domain of factual information that you can already comprehend, spaced repetition systems are pretty optimal. So, if you want to train your brain on a bunch of Spanish-to-English sentence translations, or stock market tickers, or definitions, or sample questions, you should use something like Anki.
Once you start using spaced repetition, you learn that one of the biggest limits is the card-making process. Making your own cards is time-consuming, although experience will make you much faster. Experience will also teach you what makes a better card. The 20 Rules of Formatting Knowledge pretty much spells it out for you, but I still had to make my own cards, find the sticking points, and edit them until I got a good sense for proper context and suitably short, distinct answers.
You can find other people's shared decks and skip the card making process yourself, but not without some new problems. First, you are going to learn more making the cards yourself than studying someone else's. If it's subtle material, you can make cards that fill in the gaps in your particular understanding. But if someone else studies something and makes cards to fill in their own gaps, that means that what you're studying may not cover material that you don't know you don't know. It would be nice if everyone's shared decks were a completely thorough treatment of the material, but, alas, it is not so. And the only way that you can tell is by comparing the deck and the material during your own studying.
Shared decks also just aren't all that good sometimes. Someone, I can't recall who, wrote a script to scrape the entire LW wiki and then cloze-delete the title from the article. I appreciate the idea of SRSing the LW wiki, and scripting the whole thing was undoubtedly really efficient. However, the result was usually question and answer text hundreds of words long, with tables of content in the middle, and probably too many cards of insufficient value.
Despite their problems, I think that shared decks have way more potential than their current use suggests. A well-crafted deck that gives its subject matter a thorough treatment could be more valuable than a textbook, and about as difficult to compose. But, looking at some of the best Anki decks I've come across, it will likely take more than one person to get such a deck off the ground.
Anki's .apkg files are sorta unwieldy to edit collaboratively, because there's not really a way to merge edits from multiple contributors. Luckily, we can export and import decks as text, and use version control like GitHub to do the same thing. With a GitHub-hosted collaborative deck, a team of people studying a textbook, like Thinking and Deciding, could all make flashcards as they go, add them all to the same deck, remove redundant cards, standardize the layout, tag cards appropriately, and share them with whomever else comes along. Then, anyone else who wants to study the textbook has a high-quality Anki deck to use in conjunction, and if they know how a question can be asked better, or if they find an error, or if the seventh chapter didn't really get much coverage, they can contribute to the deck, too.
This huge list of material put together by Louie Helm should be Anki-fied. Hopefully we can unite the efforts of many autodidacts and start to curate decks for each of the areas covered. Maybe a group of friends is about to work through a course on Quantum Computing or Set Theory. The rest of LW would benefit from their work making flashcards, but especially so if they leave the project open to collaboration.
So, the things needed to move forward: