omain logic. Invariably aspects of this representation would end up
> leaking into the protocol.
> 7. It was very hard to figure out what the protocol was from the code since
> the definition was embedded in byte munging code spread out over dozens of
> files.
>
> So that was
with
domain logic. Invariably aspects of this representation would end up
leaking into the protocol.
7. It was very hard to figure out what the protocol was from the code
since
the definition was embedded in byte munging code spread out over dozens of
files.
So that was definitely bad.
We consid
t;> requests.
>> 6. The same class that defined the protocol was used throughout the code.
>> So what were intended to be dumb DTOs ended up getting loaded up with
>> domain logic. Invariably aspects of this representation would end up
>> leaking into the protocol.
>&g
entation would end up
leaking into the protocol.
7. It was very hard to figure out what the protocol was from the code since
the definition was embedded in byte munging code spread out over dozens of
files.
So that was definitely bad.
We considered moving to an off-the-shelf protocol definition
ing a version change.
Guozhang
On Fri, Feb 7, 2014 at 3:23 PM, Jay Kreps wrote:
> If what you are saying is that we should define the java code and kind of
> "automatically" generate the protocol off of that then I would be against
> that as I really want the protocol definition
If what you are saying is that we should define the java code and kind of
"automatically" generate the protocol off of that then I would be against
that as I really want the protocol definition to be the source of truth. It
should be 100% obvious when you are changing the protocol.
-Ja
same class that defined the protocol was used throughout the code.
> > So what were intended to be dumb DTOs ended up getting loaded up with
> > domain logic. Invariably aspects of this representation would end up
> > leaking into the protocol.
> > 7. It was very hard to figure o
d up with
> domain logic. Invariably aspects of this representation would end up
> leaking into the protocol.
> 7. It was very hard to figure out what the protocol was from the code since
> the definition was embedded in byte munging code spread out over dozens of
> files.
>
>
col.
7. It was very hard to figure out what the protocol was from the code since
the definition was embedded in byte munging code spread out over dozens of
files.
So that was definitely bad.
We considered moving to an off-the-shelf protocol definition language like
avro or protocol buffers.
