Closed cmcnulty closed 9 years ago
Could you elaborate on the use case? Password generation is synchronous. I don't see the benefit of using a callback when you'll have the password and ready immediately upon execution. Couldn't you just pass a reference to the password generation function to accomplish what you need?
In the case of long-running password generation schemes, such as bcrypt ( https://github.com/dcodeIO/bcrypt.js ) the password generation is generally best handled as an async process because it can take several seconds to generate the password. It's true that most bcrypt javascript libraries include a synchronous option as well, but that just locks the browser until the execution is complete. In terms of SGP, it would basically mean that lines 248-263 of sgp.mobile.js would be incapsulated in their own function, which would take as an argument the completed password.
Alternatively, you could provide allow different function altogether for async which seems to be the way a lot of node packages work, but then you'd need a way to fork the decision of which function to use within the sgp.mobile.js file.
Sorry for the silence on this, still trying to wrap my head around it.
So, browsers' JS execution is still single-threaded, right? (Unless you are using Web Workers or something.) If so, what will a async/callback end up doing? Won't it still lock the browser, just a little bit later (and maybe that's enough, to let your UI code finish).
That's a very good question, and I had to search around for the answer myself, I only new that it worked! :) The answer is that, yes, JS execution is single-threaded, however node implements some functionality (which is available via browserify), which allows us to run long running processes without locking the thread. Here's a really great article about it:
http://howtonode.org/understanding-process-next-tick
which includes a chapter entitled "Interleaving execution of a CPU intensive task with other events" which is precisely how you'd want SGP to behave, if someone wanted to use a computation/memory intensive hashing algorithms (such as bcrypt or scrypt) instead of the highly efficient hashing algorithms such as MD5 and SHA.
It's much simpler than you guys debate. :)
A long-running CPU intensive task in the browser can just pause in the middle and setTimeout(next, 0)
. The browser then gets a chance to run UI, event handlers, and other code. The CPU intensive task takes slightly longer, but the thread does not lock.
This is definitely a very good point, may we continue debating it on supergenpass-lib#19?
Closing as this has been addressed in supergenpass-lib.
In order to support alternate password generation libraries, it would be nice if the SGP (and the API of supergenpass-lib) provided a mechanism for both asynchronous password generation. This would take the form of two callbacks passed into the options array of the password generator function, a progress callback and a completed callback. I already have the code in my fork, I'll split it off into a pull request whenever I get a chance, if I get an indication that such a pull would be considered.