Re: [w3c/IndexedDB] API to enumerate databases (#31)

> We should at least ensure that there is no version change transaction blocking while running this getDatabase operation, don't we?

I don't think we do...?

I'm only familiar with WebKit's implementation and in that impl this doesn't concern me. I still don't see how others can have implemented things such that they do have a huge concern here.

Here's what we do.

No matter how many unique databases are open with an active request queue, we're still piping everything through only a handful of threads all in one process.

Handling this "getAllDatabases" request will hijack the database i/o thread for the period of time it takes to tally all of the results. During that period of time literally no other database operations will happen (version change or not).

It doesn't matter if any version change transactions are in progress or not. The "getAllDatabases" operation will read the non-committed truth from disk for all in-scope databases.

If you have a database at version 1 and it has an open transaction that will be upgrading it to version 2, it will definitely and deterministically be returned as version 1 in the getAllDatabases call.

Does anybody else implement things so differently that this isn't automatically the case, or at least wouldn't be easy to make happen...?

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/w3c/IndexedDB/issues/31#issuecomment-323639751