On 31 Mar 2016, at 14:44, Eric Blake <[email protected]> wrote: > Another argument in favor of option #2: in v4, I plan to amend > NBD_REPLY_TYPE_ERROR to explicitly allow the server to optionally send > more than length 4 (and NBD_REPLY_TYPE_ERROR_OFFSET to send more than > length 12) - anything beyond the fixed fields would then be a > human-readable UTF-8 string (the same as we use for NBD_REP_ERR_* > handshake replies).
Yeah I think this is the best route. Also (hidden deep within my long comments), option #2 allows us to (optionally) send offsets for errors in NBD_CMD_WRITE as well, which is a substantial benefit. -- Alex Bligh
signature.asc
Description: Message signed with OpenPGP using GPGMail
------------------------------------------------------------------------------ Transform Data into Opportunity. Accelerate data analysis in your applications with Intel Data Analytics Acceleration Library. Click to learn more. http://pubads.g.doubleclick.net/gampad/clk?id=278785471&iu=/4140
_______________________________________________ Nbd-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/nbd-general
