Francesco wrote:
Il venerdì 09 gennaio 2009 20:58:39 Kristian Nielsen ha scritto:
Eric Day <[email protected]> writes:
Do people actually use max_allowed_packet as a limit? It seems to me
like a "hack" of how to limit large requests.
Unless someone has a good reason to keep it, I propose we remove it
in the current protocol (always use the maximum size). Also, the new
protocol draft currently does not have any limits (chunked packets).
I believe an important rationale for the max_allowed_packet is to prevent
denial-of-service, since all of the data of eg. a blob value has to be
materialised in the server (so without max_allowed_packet client can
allocate arbitrary amounts of server memory).
Still I vote for "kill the limit", if it's needed for security than can be re-
added later, left with default to unlimited and more shaped around security.
For reference last time I've seen a bug related to this was when some kde4
applications started to use mysql as a backend. The default limit of
max_allowed_packet is ridicolously low and it created random failures . You
know the best ones to find and debug.
So, the problem is not that MySQL was stupid but that the configuration
was not appropriate.
[Whether that is important for Drizzle I am not sure; I get the feeling
that the philosophy in Drizzle is to trust the client, which is fine with
me, just wanted to mention this point in case some were unaware.]
- Kristian.
Andrey
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
