Eric, On 6 Apr 2016, at 17:46, Eric Blake <eblake@...696...> wrote: > A server that sticks a super-long UTF-8 string in an error reply could > cause problems for clients. Should we have an upper bound on the length > permissible in NBD_REP_ERR_*, such as 4096, and permit clients to > disconnect if the server sends a length larger than that, so that > clients can usefully read the error message into a stack-local buffer > rather than having to heap-allocate and worry about a rogue server > sending a message as large as 2^32 bytes? +1. Similarly with export names. -- Alex Bligh
Attachment:
signature.asc
Description: Message signed with OpenPGP using GPGMail