[ot][spam][crazy] log: append-only random-access data

Undiscussed Horrific Abuse, One Victim of Many gmkarl at gmail.com
Sun Jul 10 05:33:39 PDT 2022


last stable version at
> https://lists.cpunks.org/pipermail/cypherpunks/2022-July/102291.html
>
> next step might be to change flushing so the index can be accumulated
>> during writes or mini flushes, so as to be able to limit the total size of
>> an index per flush if needed, without issue.
>>
>
> maybe this means retaining a working index as class data, and changing the
> approach for updating it so that it doesn't really solely on expanding the
> last entry but rather updates the whole thing
>
> I think this simplifies to just generating an index that includes the
previous flush, and then removing parts replaced by new writes.

So it would make sense to place writes into the index.

- generate the _next_ index on flush, referencing new writes as entries.

probably important to optimize away the approach of reiterating all leaves
>> every flush. would greatly limit write speed on older stores, and be much
>> harder to comprehend when it is old.
>>
>
> basically this could mean adding state and/or methods to internodes so
> things like their path and depth don't need to be fully recalculated
>

still looks important

>
>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: text/html
Size: 3031 bytes
Desc: not available
URL: <https://lists.cpunks.org/pipermail/cypherpunks/attachments/20220710/6a622816/attachment.txt>


More information about the cypherpunks mailing list