[erlang-questions] BERT vs protobuf in the erlang world

Steve Davis steven.charles.davis@REDACTED
Thu Aug 25 01:51:47 CEST 2011


On 8/24/2011 9:05 AM, Vincent de Phily wrote:
> On Wednesday 24 August 2011 08:37:43 Steve Davis wrote:
>> WRT to the OP, and in my simple-minded world, protobuffs always seemed
>> like a lot of hassle for not much gain;
>
> Grammar-described formats have a few advantages over self-described formats,
> such as validation, size, and speed. But self-describing formats can be great
> too. Depends on your use-case.
>
I think I would suggest that this is the same argument as "static versus 
dynamic typing". I would argue that in the cases where stronger checking 
is actually needed what's really missing is the core idea of a contract 
per UBF, rather than some standardized grammar. The distinction is 
narrow but the impact I think is quite different.

>> And if there is concern about
>> the verbosity or sending big binaries why not use gzip? Maybe there's
>> room for an extension to BERT/BERT-RPC to define the use of gzip
>> compression?
>
> I no fan of RPC frameworks, they tend to be either sinfully complicated or so
> simple that you might as well roll your own.
>
> Of course you should compress your data. But I dont think it should
> standardized in the wirelevel format. Maybe you want a compression algorythm
> that has a different speed/size tradeoff. Maybe you packets are so small or so
> tightly encoded (hello UPER !) that gziping them makes them bigger, not
> smaller.

Actually, I meant "an extension" for those situations, not as a 
requirement of the protocol... and gave gzip as the candidate in the 
spirit of keeping it all agnostic.

/s



More information about the erlang-questions mailing list