I think the general consensus has been "if the code is too long to put it into the question, it's too long to be reviewed by us" and I'd like to stick with this.
The problem with allowing to dump GitHub links is:
- It is not a resource we control, it might be gone one day (repo deletion, user deletion, GitHub blows up...) which would render the answers useless.
- It allows the users to stretch the definition of "too much code to review" by much. It's hard to argue again someone wanting us to review the whole project instead of a single file.
- If it is so long that we need to use an external service, the code is really too long, the question should be split up.
At the moment the code that gets posted has a reasonable length most of the time, it's limited to some functions, maybe two or three classes and all in all fits very well. Posting more code would for sure limit the audience of the question ("I'm not going to review all this crap"), so it would be better if the user splits it up and presents it in different questions that can stand on their own (so no "Authentication on website - Part x of y").
If we look at it from the users perspective, being able to dump all code at once on us is a huge advantage. But if we look at it from the perspective of everyone else, having short, concise and self-contained pieces of code, does not only allow the answers to be more specialized, but it will most likely also attract more reviewers (because they ain't need to spent that much time on it) and does also allow us to use the questions for reference purposes in the future.
In short, I think I'd sum it up like this:
First step [for a code review here] is to only show that part of the project that you want to be reviewed, if you already fail at reasonably extracting that part so that it can stand on it's own, you're already doing it wrong.