[PATCH 0/2] CNFSv4: move to 4K blocksize

["Miquel van Smoorenburg"]

In article <87y6z6bc6x.fsf at windlord.stanford.edu>,
Russ Allbery  <rra at stanford.edu> wrote:
>list-inn-workers at news.cistron.nl ("Miquel van Smoorenburg" ) writes:
>
>> This introduces a new CNFS version in the CNFS header, version 4.
>
>If we're going to introduce a new version of the CNFS format, maybe we
>could deal with the other long-standing problem at the same time, namely
>that the CNFS file format depends on the configuration parameters for the
>INN build?  If the CNFSv4 file format standardized uint32_t and uint64_t
>values in network byte order, that would let us move CNFS buffers between
>hosts and between largefile and non-largefile builds.

Good point. The CNFS header is ok.

The values in the article header are stored in network byte order,
but 'long' can be 32 or 64 bits (and so can time_t).
I think we should make 'size' an uint32_t, 'arrived' an int32_t+uint32_t
(upper and lower 32 bits) and 'class' an uint32_t as well.

The other thing that needs to be fixed is the bitmap code in
CNFSUsedBlock().

Is it a good idea to remove the OLD_CNFS code ? It's not compiled
in by default, and I doubt anyone still uses it. In that case I
can recycle the 'oldCNFSARTHEADER' name :)

Mike.
-- 
The From: and Reply-To: addresses are internal news2mail gateway addresses.
Reply to the list or to "Miquel van Smoorenburg" <miquels at cistron.nl>