Hi Daniel,
I'll fly back to Europe tomorrow & the day after.
Once there, I'll take time to answer your questions
and review your patch (elsethread).
-- Stefan^2.
On Mon, May 18, 2015 at 1:20 AM, Daniel Shahaf <d.s_at_daniel.shahaf.name>
wrote:
> Good morning Stefan,
>
> A couple of questions on the fsfs7 indexes doc:
>
> [[[
> --- subversion/libsvn_fs_fs/structure-indexes
> +++ subversion/libsvn_fs_fs/structure-indexes
> @@ -18,6 +18,9 @@ a simple concatenation of runtime structs and as such,
> an implementation
> detail subject to change. A proto index basically aggregates all the
> information that must later be transformed into the final index.
>
> +### "Subject to change?" The format must be stable & documented since
> +### "begin txn, modify txn, reboot, upgrade svn, commit txn" should work.
> +
>
> General design concerns
> -----------------------
> @@ -346,7 +349,10 @@ For performance reasons we use a modified version:
> * combine the big endian representation of these checksums plus the
> remnant of the original stream into a 12 to 15 byte long intermediate
>
> [i0 .. iK], 12 <= K+1 <= 15
>
> +### "combine" is unclear. Combine how? Concatenate? Interleave? Add?
> Multiply?
> +
> * FNV checksum = fnv_1a([i0 .. iK]) in big endian representation
>
> +### Why do we checksum the output of a checksum algorithm? (Bytes i0..iK
> are themselves FNV output)
> ]]]
>
> Also, I suggest to change "zero to many" to "zero or more" to avoid
> confusion
> with the term "one-to-many" of relational databases.
>
> Cheers,
>
> Daniel
>
Received on 2015-05-18 02:59:17 CEST