Tweaks

[jauntywunderkind]

Hello. I've been interested in reading my own snss & indexeddb data out, to better track my own activity over time.

I found a couple modest tweaks that I needed to make use of this library. Please let me know if there's anything I can do to help get some or all of these changes merged in.

[cclgroupltd]

I think the Crypto/Cryptodome difference might just be a local environment thing? Did you install the packages from requirements.txt because there are two versions of pycryptodome and the only difference is the import name.

I'd also like to leave win32crypt in as we're doing most of our research on Windows here.

[cclgroupltd]

I favour specificity of the caller passing the blob path over this default behaviour I think.

[jauntywunderkind]

What does that help, what's the advantage? I'm trying to understand. Is there something else we can do?

This seems like easy to derive data that WrappedIndexDB can calculate. Why would we burden callers with doing this scut work, when it is easy to derive like this as a default parameter?

We could eliminate the parameter entirely, removing likely never needed flexibility?

[cclgroupltd]

I'll counter this by saying: I used this flexibility last week when I had a blob folder containing corrupt data recovered from a disk image. There might also be situations where the blob folder straight up doesn't exist and as a caller I'd want to know about it early rather than deep within another call. Explicit is better than implicit, etc.

I think it's worth bearing in in that the original reason for this module to exist is for digital forensics and research where this flexibility is sometimes desirable and where precision and traceability is always going to trump the minor inconvenience of checking for a blob folder's presence and adding it in my eyes.

[jauntywunderkind]

Thanks for the fast comments, much appreciated! I've tried to bring things closer in line to what you've said.

I really like that we can avoid doing the get_blob_info lookup twice now, which was kind of expensive ("goodness me this is a slow way of doing things,").

I don't really know much about python packaging/library situations to be honest. On Debian my system had all these libraries available (except win32), except the the system python3-pycryptodome gets installed installed as Cryptodome not Crypto. I wouldn't know what to do to try to do to create some kind of virtual env where I could safely install python packages without affecting my whole system. The changes I made are reverted out, so there's both a record of what someone needs to do, while not affecting the current behavior. In #15 (comment) I saw mention of a try import which I didn't know about; in a future PR I might try using that to make audit better able to fallback to Linux runnability, but I don't expect to do that exploration in this PR.