On 21/02/16 21:15, Heikki Linnakangas wrote:
> On 19/02/16 10:10, Álvaro Hernández Tortosa wrote:
>> Oleg and I discussed recently that a really good addition to a
>> GSoC
>> item would be to study whether it's convenient to have a binary
>> serialization format for jsonb over the wire. Some argue this should be
>> benchmarked first. So the scope for this project would be to benchmark
>> and analyze the potential improvements and then agree on which format
>> jsonb could be serialized to (apart from the current on-disk format,
>> there are many json or nested k-v formats that could be used for sending
>> over the wire).
>
> Seems a bit risky for a GSoC project. We don't know if a different
> serialization format will be a win,
Over the current serialization (text) is hard to believe there will
be no wins.
> or whether we want to do it in the end, until the benchmarking is
> done. It's also not clear what we're trying to achieve with the
> serialization format: smaller on-the-wire size, faster serialization
> in the server, faster parsing in the client, or what?
Probably all of them (it would be ideal if it could be selectable).
Some may favor small on-the-wire size (which can be significant with
several serialization formats) or faster decoding (de-serialization
takes a significant execution time). Of course, all this should be
tested and benchmarked before, but we're not alone here.
This is a significant request from many, at least from the Java
users, where it has been discussed many times. Specially if wire format
adheres to one well-known (or even Standard) format, so that the
receiving side and the drivers could expose an API based on that format
--one of the other big pains today in this side.
I think it fits very well for a GSoC! :)
Álvaro
--
Álvaro Hernández Tortosa
-----------
8Kdata