how much memory is used per item / false positive percentage?

[kolinfluence]

1. lower false positive rates = higher memory used. is there a vague guide on the number of bytes?

    filter := bloom.NewWithEstimates(1000000, 0.01) 

2. for a cdn deployment with 100 million items, what's a good number for the bloom item count and false positive rate?
   also how much memory is needed for this bloom filter?

3. i was thinking if it's using a lot of memory, would it be advisable to use roaring bitmaps as "replacement" to bloom filter as suggested by chatgpt here:
   Regarding memory use of maximum size and removal of bit number RoaringBitmap/roaring#388