← Back to team overview

maria-discuss team mailing list archive

Re: MariaDB Server 10.3 notes


Antony's work includes a client library change that lets perl/Java/or whatever run SQL on the THD.  The parser was made reentrant, you connect with null credential and get current THD.

Sent from my iPhone

> On Oct 13, 2016, at 1:01 PM, Andrew Hutchings <andrew@xxxxxxxxxxxxxxx> wrote:
>> On 13/10/16 16:53, Vladislav Vaintroub wrote:
>>> On 13.10.2016 17:26, Andrew Hutchings wrote:
>>> Whilst I agree it would be a nice feature and something I've heard for
>>> years, the implementation would likely be more complex than it seems.
>>> For starters you have to consider the possibility of a procedure
>>> crashing and taking the whole daemon with it. You would likely need to
>>> fork a small worker process pool and have some kind of shared memory
>>> or socket communications for safety. In addition the implementation
>>> would have to be extremely careful not to add any potential security
>>> hole due to a zero-day in PHP or some bad input filtering for example.
>>> I'm not saying it is impossible, but it will likely be a lot of work
>>> to get right and the APIs would need to be carefully thought out.
>>> So the question becomes: is it worth spending time developing this
>>> over another feature? Or is it something that could be better
>>> implemented safely in another layer, such as in a database proxy?
>>> Kind Regards
>>> Andrew
>> I think crashing is much easier if one runs native C UDFs, or any kind
>> of natively compiled plugins.
> I don't know, I've seen some horrid things happen with interpreted languages before.
>> For external languages I think an important aspect is often overlooked.
>> Namely, how to would access data. You can create a new JDBC connection,
>> but it is likely not what you want to do. You just want to run a SELECT
>> from the same session you're in, just like you do in SQL stored
>> procedures. Exactly this requires a large amount of work in the
>> connectors. It is not  just running a JVM  inside the server.  There is
>> no API for data access from the running session inside the server,  not
>> even for C.
> I was envisioning a tighter coupling than using connectors in the above. Opening up the thd to a new process via a new API and blocking the client whilst doing so. I hadn't really thought though the mechanics of it too much though. But you are right, either way it is a lot of work.
> Again, I think this kind of thing is better suited to the application or proxy layers. Especially with things such as MicroServices and Serverless around which are designed to break up the logic like this (I'm not a fan of the buzzword though).
> Kind Regards
> -- 
> Andrew Hutchings - LinuxJedi - http://www.linuxjedi.co.uk/
> _______________________________________________
> Mailing list: https://launchpad.net/~maria-discuss
> Post to     : maria-discuss@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~maria-discuss
> More help   : https://help.launchpad.net/ListHelp

Follow ups